matlab @ Image pro

Image Processing Toolbox ® For Use with MATLAB Computa tion Visua liza tion Progra m ming User’s Guide Version 2.1 www.MatlabSite.com How to Contact The M athW orks: ☎ 508-647-7000 Phone 508-647-7001 Fax The MathWorks, Inc. 24 Prime Park Way Natick, MA 01760-1500 Mail http://www.mathworks.com Web Anonymous FTP server Newsgroup PHONE FAX ✉ MAIL INTERNET ftp.mathworks.com comp.soft-sys.matlab @ support@mathworks.com suggest@mathworks.com bugs@mathworks.com doc@mathworks.com subscribe@mathworks.com service@mathworks.com info@mathworks.com Technical support Product enhancement suggestions Bug reports Documentation error reports Subscribing user registration Order status, license renewals, passcodes Sales, pricing, and general information Image Processing Toolbox User’s Guide  COPYRIGHT 1993 - 1998 by The MathWorks, Inc. All Rights Reserved. The software described in this document is furnished under a license agreement. The software may be used or copied only under the terms of the license agreement. No part of this manual may be photocopied or reproduced in any form without prior written consent from The MathWorks, Inc. U.S. GOVERNMENT: If Licensee is acquiring the Programs on behalf of any unit or agency of the U.S. Government, the following shall apply: (a) For units of the Department of Defense: the Government shall have only the rights specified in the license under which the commercial computer software or commercial software documentation was obtained, as set forth in subparagraph (a) of the Rights in Commercial Computer Software or Commercial Software Documentation Clause at DFARS 227.7202-3, therefore the rights set forth herein shall apply; and (b) For any other unit or agency: NOTICE: Notwithstanding any other lease or license agreement that may pertain to, or accompany the delivery of, the computer software and accompanying documentation, the rights of the Government regarding its use, reproduction, and disclosure are as set forth in Clause 52.227-19 (c)(2) of the FAR. MATLAB, Simulink, Handle Graphics, and Real-Time Workshop are registered trademarks and Stateflow and Target Language Compiler are trademarks of The MathWorks, Inc. Other product or brand names are trademarks or registered trademarks of their respective holders. Printing History: August 1993 May 1997 January 1998 First printing Version 1 Second printing Version 2 Revised for MATLAB 5.2 (online version) www.MatlabSite.com Ima ge Credits moon Copyright Michael Myers. Used with permission. cameraman Copyright Massachusetts Institute of Technology. Used with permission. trees Trees with a View, watercolor and ink on paper, copyright Susan Cohen. Used with permission. forest Photograph of Carmanah Ancient Forest, British Columbia, Canada, courtesy of Susan Cohen. circuit Micrograph of 16-bit A/D converter circuit, courtesy of Steve Decker and Shujaat Nadeem, MIT, 1993. m83 M83 spiral galaxy astronomical image courtesy of Anglo-Australian Observatory, photography by David Malin. alumgrns bacteria blood1 bonemarr circles circlesm debye1 enamel flowers ic ngc4024l ngc4024m ngc4024s rice saturn shot1 testpat1 testpat2 text tire Copyright J. C. Russ, The Image Processing Handbook, Second Edition, 1994, CRC Press, Boca Raton, ISBN 0-8493-2516-1. Used with permission. www.MatlabSite.com www.MatlabSite.com Contents Before You Begin What is the Image Processing Toolbox? . . . . . . . . . . . . . . . . . . . . . . . x New Features in Version 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x Installing the Toolbox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi About This Manual . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii Typographical Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii Introduction 1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-2 Images in MATLAB and the Image Processing Toolbox . . . . 1-3 Data Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-3 Image Types in the Toolbox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-5 Indexed Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-5 Intensity Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-6 Binary Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-7 RGB Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-8 Multiframe Image Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-9 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-10 Working with Image Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Reading and Writing Images . . . . . . . . . . . . . . . . . . . . . . . . . . Converting Images to Other Types . . . . . . . . . . . . . . . . . . . . . . Color Space Conversions . . . . . . . . . . . . . . . . . . . . . . . . . . . . Working with uint8 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Converting Between Data Types . . . . . . . . . . . . . . . . . . . . . Turning the Logical Flag on or off . . . . . . . . . . . . . . . . . . . . www.MatlabSite.com 1-11 1-11 1-11 1-13 1-13 1-14 1-15 i Coordinate Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pixel Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Spatial Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Using a Nondefault Spatial Coordinate System . . . . . . . . . 1-16 1-16 1-17 1-18 Displaying Images 2 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-2 Standard Display Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-3 Displaying Images with imshow . . . . . . . . . . . . . . . . . . . . . . . . . 2-3 Preferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-3 The truesize Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-4 Displaying Indexed Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-5 Displaying Intensity Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-5 Displaying Binary Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-6 Changing the Display Colors . . . . . . . . . . . . . . . . . . . . . . . . . 2-7 Displaying RGB Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-7 Displaying Images in Graphics Files . . . . . . . . . . . . . . . . . . . . . 2-8 Displaying Nonimage Data as an Intensity Image . . . . . . . . . . 2-8 Special Display Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Adding a Colorbar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Displaying Multiframe Image Arrays . . . . . . . . . . . . . . . . . . . . Displaying Frames Individually . . . . . . . . . . . . . . . . . . . . . . Displaying All Frames at Once . . . . . . . . . . . . . . . . . . . . . . . Converting the Array to a Movie . . . . . . . . . . . . . . . . . . . . . Displaying Multiple Images . . . . . . . . . . . . . . . . . . . . . . . . . . . Displaying Each Image in a Separate Figure . . . . . . . . . . . Displaying Multiple Images in the Same Figure . . . . . . . . . Zooming in on a Region of an Image . . . . . . . . . . . . . . . . . . . . . Texture Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-10 2-10 2-11 2-11 2-12 2-13 2-14 2-14 2-15 2-16 2-17 Printing Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-19 ii Contents www.MatlabSite.com Geometric Operations 3 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-2 Interpolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-3 Image Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-4 Image Resizing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-5 Image Rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-6 Image Cropping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-7 Neighborhood and Block Operations 4 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-2 Types of Block Processing Operations . . . . . . . . . . . . . . . . . . . . 4-2 Sliding Neighborhood Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-4 Padding of Borders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-5 Linear and Nonlinear Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . 4-5 Distinct Block Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-8 Overlap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-9 Column Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sliding Neighborhoods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Distinct Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Restrictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . www.MatlabSite.com 4-11 4-11 4-12 4-14 iii Linear Filtering and Filter Design 5 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-2 Linear Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-3 Convolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-3 Rotating the Convolution Kernel . . . . . . . . . . . . . . . . . . . . . . 5-4 Determining the Center Pixel . . . . . . . . . . . . . . . . . . . . . . . . . 5-4 Applying the Computational Molecule . . . . . . . . . . . . . . . . . . 5-4 Padding of Borders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-5 The filter2 Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-7 Separability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-8 Determining Separability . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-9 Higher-Dimensional Convolution . . . . . . . . . . . . . . . . . . . . . . . . 5-9 Using Predefined Filter Types . . . . . . . . . . . . . . . . . . . . . . . . . . 5-9 Filter Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . FIR Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Frequency Transformation Method . . . . . . . . . . . . . . . . . . . . . Frequency Sampling Method . . . . . . . . . . . . . . . . . . . . . . . . . . Windowing Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Creating the Desired Frequency Response Matrix . . . . . . . . . Computing the Frequency Response of a Filter . . . . . . . . . . . . 5-13 5-13 5-14 5-15 5-16 5-18 5-19 Transforms 6 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-2 Fourier Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-3 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-3 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-4 The Discrete Fourier Transform . . . . . . . . . . . . . . . . . . . . . . . . . 6-8 Relationship to the Fourier Transform . . . . . . . . . . . . . . . . . 6-8 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-9 iv Contents www.MatlabSite.com Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Frequency Response of Linear Filters . . . . . . . . . . . . . . . . . Fast Convolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Locating Image Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-11 6-11 6-12 6-13 Discrete Cosine Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-15 The DCT Transform Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-16 The DCT and Image Compression . . . . . . . . . . . . . . . . . . . . . . 6-17 Radon Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-19 Using the Radon Transform to Detect Lines . . . . . . . . . . . . . . 6-22 The Inverse Radon Transform . . . . . . . . . . . . . . . . . . . . . . . . . 6-25 Analyzing and Enhancing Images 7 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-2 Pixel Values and Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-3 Pixel Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-3 Intensity Profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-4 Image Contours . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-7 Image Histogram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-8 Summary Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-9 Feature Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-9 Image Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-10 Edge Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-10 Quadtree Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-11 www.MatlabSite.com v Image Enhancement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Intensity Adjustment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gamma Correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Histogram Equalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . Noise Removal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Linear Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Median Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Adaptive Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-14 7-14 7-16 7-18 7-20 7-21 7-21 7-23 Binary Image Operations 8 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-2 Neighborhoods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-2 Padding of Borders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-2 Displaying Binary Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-3 Morphological Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-4 Dilation and Erosion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-4 Related Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-7 Predefined Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-8 Object-Based Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4- and 8-Connected Neighborhoods . . . . . . . . . . . . . . . . . . . . . . Perimeter Determination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Flood Fill . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Connected-Components Labeling . . . . . . . . . . . . . . . . . . . . . . . Object Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-10 8-10 8-12 8-13 8-15 8-16 Feature Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-18 Image Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-18 Euler Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-19 Lookup-Table Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-20 vi Contents www.MatlabSite.com Region-Based Processing 9 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-2 Specifying a Region of Interest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-3 Selecting a Polygon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-3 Other Selection Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-5 Filtering a Region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-6 Filling a Region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-8 Color 10 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-2 Working with Different Color Depths . . . . . . . . . . . . . . . . . . . . . 10-3 Reducing the Number of Colors in an Image . . . . . . . . . . . . . . Using rgb2ind . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Colormap Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Using imapprox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dithering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-5 10-5 10-6 10-7 10-7 10-8 Converting to Other Color Spaces . . . . . . . . . . . . . . . . . . . . . . . . . 10-9 NTSC Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-9 YCbCr Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-10 HSV Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-10 Reference 11 www.MatlabSite.com vii viii Contents www.MatlabSite.com Before You Begin What is the Image Processing Toolbox? . . . . . . . . . x New Features in Version 2.1 . . . . . . . . . . . . . . . x Installing the Toolbox . . . . . . . . . . . . . . . . . . xi About This Manual . . . . . . . . . . . . . . . . . .xii Typographical Conventions . . . . . . . . . . . . . . xiii www.MatlabSite.com Before You Begin W ha t is the Ima ge Processing Toolbox ? The Image Processing Toolbox is a collection of functions that extend the capability of the MATLAB® numeric computing environment. The toolbox supports a wide range of image processing operations, including: • Geometric operations • Neighborhood and block operations • Linear filtering and filter design • Transforms • Image analysis and enhancement • Binary image operations • Region of interest operations Many of the toolbox functions are MATLAB M-files, series of MATLAB statements that implement specialized image processing algorithms. You can view the MATLAB code for these functions using the statement: type function_name You can extend the capabilities of the Image Processing Toolbox by writing your own M-files, or by using the toolbox in combination with with other toolboxes, such as the Signal Processing Toolbox and the Wavelet Toolbox. N ew Fea tures in Version 2 .1 Version 2.1 of the Image Processing Toolbox offers a number of advances over versions 1 and 2. For information about the differences between versions 1 and 2, and versions 2 and 2.1, type this command at the MATLAB prompt: helpwin images/Readme For a list of all of the functions in the Image Processing Toolbox, type this command: helpwin images/Contents x www.MatlabSite.com W hat is the Image Processing Toolbox? Insta lling the Toolbox To determine if the Image Processing Toolbox is installed on your system, type this command at the MATLAB prompt: ver When you enter this command, MATLAB displays information about the version of MATLAB you are running, including a list of all toolboxes installed on your system and their version numbers. For information about installing the toolbox, see the MATLAB Installation Guide for your platform. www.MatlabSite.com xi Before You Begin About This M a nua l This manual has three main parts: • Chapters 1 and 2 discuss working with image data and displaying images in MATLAB and the Image Processing Toolbox. • Chapters 3 to 10 provide in-depth discussion of the concepts behind the software. Each chapter covers a different topic in image processing. For example, Chapter 5 discusses linear filtering, and Chapter 9 discusses binary image operations. Each chapter provides numerous examples that apply the toolbox to representative image processing tasks. • Chapter 11 gives a detailed reference description of each toolbox function. Reference descriptions include a synopsis of the function’s syntax, as well as a complete explanation of options. Many reference descriptions also include examples, a description of the function’s algorithm, and references to additional reading material. All users should read Chapters 1 and 2. Less experienced users will find Chapters 3 to 10 a valuable introduction to image processing, while more experienced users may prefer to use Chapter 11. xii www.MatlabSite.com About This M anual Ty pogra phica l Conventions To Indica te... This M a nua l Uses... Ex a m ple Example code Monospace type To assign the value 5 to A, enter: A = 5 MATLAB output Monospace type MATLAB responds with: A = 5 Function names Monospace type The cos function finds the cosine of each array element. New terms Italics An array is an ordered collection of information. Keys Boldface with an initial capital letter Press the Return key. Menu names, items, and GUI controls Boldface with an initial capital letter Choose the Print option from the File menu. Mathematical expressions Variables in italics. Functions, operators, and constants in standard type This vector represents the polynomial: www.MatlabSite.com p = x2 + 2x + 3 xiii Before You Begin xiv www.MatlabSite.com 1 Introduction Overview . . . . . . . . . . . . . . . . . . . . . 1-2 Images in MATLAB and the Image Processing Toolbox 1-3 Data Types . . . . . . . . . . . . . . . . . . . . . 1-3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-5 1-5 1-6 1-7 1-8 1-9 Working with Image Data . . . Reading and Writing Images . . . Converting Images to Other Types Working with uint8 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-11 1-11 1-11 1-13 Image Types in the Toolbox Indexed Images . . . . . . Intensity Images . . . . . . Binary Images . . . . . . . RGB Images . . . . . . . . Multiframe Image Arrays . . . . . . . . Coordinate Systems . . . . . . . . . . . . . . . . 1-16 Pixel Coordinates . . . . . . . . . . . . . . . . . . 1-16 Spatial Coordinates . . . . . . . . . . . . . . . . . . 1-17 www.MatlabSite.com 1 Introduction O verview This chapter introduces you to the fundamentals of image processing using MATLAB and the Image Processing Toolbox. It describes the types of images supported, and how MATLAB represents them. It also explains the basics of working with image data and coordinate systems. 1-2 www.MatlabSite.com Images in M ATLAB and the Image Processing Toolbox Ima ges in M ATLAB a nd the Ima ge Processing Toolbox The basic data structure in MATLAB is the array, an ordered set of real or complex elements. This object is naturally suited to the representation of images, real-valued, ordered sets of color or intensity data. (MATLAB does not support complex-valued images.) MATLAB stores most images as two-dimensional arrays (i.e., matrices), in which each element of the matrix corresponds to a single pixel in the displayed image. (Pixel is derived from picture element and usually denotes a single dot on a computer display.) For example, an image composed of 200 rows and 300 columns of different colored dots would be stored in MATLAB as a 200-by-300 matrix. This convention makes working with images in MATLAB similar to working with any other type of matrix data, and makes the full power of MATLAB available for image processing applications. For example, you can select a single pixel from an image matrix using normal matrix subscripting: I(2,15) This command returns the value of the pixel at row 2, column 15 of the image I. Da ta Ty pes By default, MATLAB stores most data in arrays of class double. The data in these arrays is stored as double precision (64-bit) floating-point numbers. All of MATLAB’s functions and capabilities work with these arrays. For image processing, however, this data representation is not always ideal. The number of pixels in an image may be very large; for example, a 1000-by-1000 image has a million pixels. Since each pixel is represented by at least one array element, this image would require about 8 megabytes of memory. In order to reduce memory requirements, MATLAB supports storing image data in arrays of class uint8. The data in these arrays is stored as 8-bit unsigned integers. Data stored in uint8 arrays requires one eighth as much memory as data in double arrays. Because the types of values that can be stored in uint8 arrays and double arrays differ, the Image Processing Toolbox uses different conventions for interpreting the values in these arrays. (Noninteger values cannot be stored in www.MatlabSite.com 1-3 1 Introduction uint8 arrays, for example, but they can be stored in double arrays.) The next section discusses how the toolbox interprets image data, depending on the class of the data array. In addition to differences in the types of data values they store, uint8 arrays and double arrays differ in the operations that MATLAB supports. See page 1-13 for information about the operations MATLAB supports for uint8 arrays. 1-4 www.MatlabSite.com Image Types in the Toolbox Ima ge Types in the Toolbox The Image Processing Toolbox supports four basic types of images: • Indexed images • Intensity images • Binary images • RGB images This section discusses how MATLAB and the Image Processing Toolbox represent each of these image types. Index ed Im a ges An indexed image consists of two arrays, an image matrix and a colormap. The colormap is an ordered set of values that represent the colors in the image. For each image pixel, the image matrix contains a value that is an index into the colormap. The colormap is an m-by-3 matrix of class double. Each row of the colormap matrix specifies the red, green, and blue (RGB) values for a single color: color = [R G B] R, G, and B are real scalars that range from 0 (black) to 1.0 (full intensity). www.MatlabSite.com 1-5 1 Introduction The figure below illustrates the structure of an indexed image. The pixels in the image are represented by integers, which are pointers (indices) to color values stored in the colormap. 75 10 12 21 40 53 53 75 14 17 21 21 53 53 75 8 5 8 10 30 15 51 15 18 31 31 18 16 56 31 18 31 31 31 31 0 0.0627 0.2902 0 0.2902 0.3882 0.4510 0.2588 0 0.0627 0.0314 0 0.0627 0.0314 0.0627 0.1608 . . . 0 0.0314 0 1.0000 0.0627 0.0941 0 0.0627 The relationship between the values in the image matrix and the colormap depends on whether the image matrix is of class double or uint8. If the image matrix is of class double, the value 1 points to the first row in the colormap, the value 2 points to the second row, and so on. If the image matrix is of class uint8, there is an offset; the value 0 points to the first row in the colormap, the value 1 points to the second row, and so on. The uint8 convention is also used in graphics file formats, and enables 8-bit indexed images to support up to 256 colors. In the image above, the image matrix is of class double, so there is no offset. For example, the value 5 points to the fifth row of the colormap. Intensity Im a ges MATLAB stores an intensity image as a single matrix, with each element of the matrix corresponding to one image pixel. The matrix can be of class double, in which case it contains values in the range [0,1], or of class uint8, in which case the data range is [0,255]. The elements in the intensity matrix represent various intensities, or gray levels, where the intensity 0 represents black and the intensity 1 (or 255) represents full intensity, or white. 1-6 www.MatlabSite.com Image Types in the Toolbox This figure depicts an intensity image of class double. 0.5342 0.5342 0.5342 0.4308 0.4510 0.2051 0.2251 0.1789 0.2483 0.3344 0.2157 0.2563 0.1307 0.2624 0.2624 0.2826 0.2826 0.1789 0.3344 0.3344 0.3822 0.2826 0.2051 0.3344 0.3344 0.4391 0.4391 0.3256 0.2624 0.3344 0.4391 0.4391 0.2483 0.2549 0.3344 Bina r y Im a ges In a binary image, each pixel assumes one of only two discrete values. Essentially, these two values correspond to on and off. A binary image is stored as a two-dimensional matrix of 0’s (off pixels) and 1’s (on pixels). A binary image can be considered a special kind of intensity image, containing only black and white. Other interpretations are possible, however; you can also think of a binary image as an indexed image with only two colors. A binary image can be stored in an array of class double or uint8. However, a uint8 array is preferable, because it uses far less memory. In the Image Processing Toolbox, any function that returns a binary image returns it as a uint8 logical array. The toolbox uses the presence of the logical flag to signify that the data range is [0,1]. (If the logical flag is off, the toolbox assumes the data range is [0,255].) www.MatlabSite.com 1-7 1 Introduction This figure shows an example of a binary image. 1 0 0 0 0 0 1 0 1 0 0 0 0 0 1 0 1 0 0 0 0 0 1 0 0 0 1 0 0 1 0 0 0 0 0 1 0 1 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 RGB Im a ges Like an indexed image, an RGB image represents each pixel color as a set of three values, representing the red, green, and blue intensities that make up the color. Unlike an indexed image, however, these intensity values are stored directly in the image array, not indirectly in a colormap. In MATLAB, the red, green, and blue components of an RGB image reside in a single m-by-n-by-3 array. m and n are the numbers of rows and columns of pixels in the image, and the third dimension consists of three planes, containing red, green, and blue intensity values. For each pixel in the image, the red, green, and blue elements combine to create the pixel’s actual color. For example, to determine the color of the pixel (112,86), look at the RGB triplet stored in (112,86,1:3). Suppose (112,86,1) contains the value 0.1238, (112,86,2) contains 0.9874, and (112,86,3) contains 0.2543. The color for the pixel at (112,86) is: 0.1238 0.9874 0.2543 1-8 www.MatlabSite.com Image Types in the Toolbox An RGB array can be of class double, in which case it contains values in the range [0,1], or of class uint8, in which case the data range is [0,255]. The figure below shows an RGB image of class double: 0.5804 0.2235 0.1294 0.2902 Blue 0.4196 0.4824 0.4824 0.5804 0.2902 0.0627 0.2902 0.2902 0.4824 0.4824 0.5804 0.0627 0.0627 0.0627 0.2235 0.2588 0.2588 0.5176 0.2588 0.0627 0.0941 0.0941 0.0627 0.0627 0.5176 0.1922 0.0627 0.1294 0.1922 0.2588 0.2588 Green 0.4510 0.0941 0.0627 0.0941 0.0941 0.0941 0.0941 0.5176 0.1294 0.1608 0.1294 0.1294 0.2588 0.2588 0.5176 0.1608 0.0627 0.1608 0.1922 0.2588 0.2588 0.4196 0.2588 0.3529 0.4196 0.4196 0.3529 0.2902 0.5490 0.2235 0.5490 0.5804 0.7412 0.7765 0.7765 Red 0.4510 0.4196 0.3529 0.4196 0.4196 0.4196 0.4196 0.5490 0.3882 0.5176 0.5804 0.5804 0.7765 0.7765 0.5490 0.2588 0.2902 0.2588 0.2235 0.4824 0.2235 0.4196 0.2235 0.1608 0.2588 0.2588 0.1608 0.2588 0.4510 0.2588 0.1608 0.2588 0.2588 0.2588 0.2588 M ultifra m e Im a ge Arra y s For some applications, you may need to work with collections of images related by time or view, such as magnetic resonance imaging (MRI) slices or movie frames. The Image Processing Toolbox provides support for storing multiple images in the same array. Each separate image is called a frame. If an array holds multiple frames, they are concatenated along the fourth dimension. For example, an array with five 400-by-300 RGB images would be 400-by-300-by-3-by-5. A similar multiframe intensity or indexed image would be 400-by-300-by-1-by-5. www.MatlabSite.com 1-9 1 Introduction For example, if you have a group of images A1, A2, A3, A4, and A5, you can store them in a single array using: A = cat(4,A1,A2,A3,A4,A5) You can also extract frames from a multiframe image. For example, if you have a multiframe image MULTI, this command extracts the third frame: FRM3 = MULTI(:,:,:,3) Note that in a multiframe image array, each image must be the same size and have the same number of planes. In a multiframe indexed image, each image must also use the same colormap. Limitations Many of the functions in the toolbox operate only on the first two or first three dimensions. You can still use four-dimensional arrays with these functions, but you must process each frame individually. For example, this call displays the seventh frame in the array MULTI: imshow(MULTI(:,:,:,7)) If you pass an array to a function and the array has more dimensions than the function is designed to operate on, your results may be unpredictable. In some cases, the function will simply process the first frame of the array, but in other cases the operation will not produce meaningful results. See the reference entries in Chapter 11 for information about how individual functions work with the dimensions of an image array. For more information about displaying the images in a multiframe array, see Chapter 2. 1-10 www.MatlabSite.com W orking w ith Image Data W ork ing w ith Ima ge Da ta This section discusses ways of working with the data arrays that represent images, including: • Reading in image data from files, and writing image data out to files • Converting images to other image types • Working with uint8 arrays in MATLAB and the Image Processing Toolbox Rea ding a nd W riting Im a ges You can use the MATLAB imread function to read image data from files. imread can read these graphics file formats: • BMP • HDF • JPEG • PCX • TIFF • XWD To write image data from MATLAB to a file, use the imwrite function. imwrite can write the same file formats that imread reads. In addition, you can use the imfinfo function to return information about the image data in a file. See the reference entries for imread, imwrite, and imfinfo in Chapter 11 for more information about these functions. Conver ting Im a ges to O ther Ty pes For certain operations, it is helpful to convert an image to a different image type. For example, if you want to filter a color image that is stored as an indexed image, you should first convert it to RGB format. When you apply the filter to the RGB image, MATLAB filters the intensity values in the image, as is appropriate. If you attempt to filter the indexed image, MATLAB simply applies the filter to the indices in the indexed image matrix, and the results may not be meaningful. www.MatlabSite.com 1-11 1 Introduction The Image Processing Toolbox provides several functions that enable you to convert any image to another image type. These functions have mnemonic names; for example, ind2gray converts an indexed image to a grayscale intensity format. Note that when you convert an image from one format to another, the resulting image may look different from the original. For example, if you convert a color indexed image to an intensity image, the resulting image is grayscale, not color. For more information about how these functions work, see their reference entries in Chapter 11. The table below summarizes these image conversion functions: 1-12 Function Purpose dither Create a binary image from a grayscale intensity image by dithering; create an indexed image from an RGB image by dithering gray2ind Create an indexed image from a grayscale intensity image grayslice Create an indexed image from a grayscale intensity image by thresholding im2bw Create a binary image from an intensity image, indexed image, or RGB image, based on a luminance threshold ind2gray Create a grayscale intensity image from an indexed image ind2rgb Create an RGB image from an indexed image mat2gray Create a grayscale intensity image from data in a matrix, by scaling the data rgb2gray Create a grayscale intensity image from an RGB image rgb2ind Create an indexed image from an RGB image www.MatlabSite.com W orking w ith Image Data You can also perform certain conversions just using MATLAB syntax. For example, you can convert an intensity image to RGB format by concatenating three copies of the original matrix along the third dimension: RGB = cat(3,I,I,I); The resulting RGB image has identical matrices for the red, green, and blue planes, so the image displays as shades of gray. In addition to these standard conversion tools, there are some functions that return a different image type as part of the operation they perform. For example, the region of interest routines each return a binary image that you can use to mask an indexed or intensity image for filtering or for other operations. Color Space Conversions The Image Processing Toolbox represents colors as RGB values, either directly (in an RGB image) or indirectly (in an indexed image). However, there are other methods for representing colors. For example, a color can be represented by its hue, saturation, and value components (HSV). Different methods for representing colors are called color spaces. The toolbox provides a set of routines for converting between RGB and other color spaces. The image processing functions themselves assume all color data is RGB, but you can process an image that uses a different color space by first converting it to RGB, and then converting the processed image back to the original color space. For more information about color space conversion routines, see Chapter 10. W ork ing w ith uint8 Da ta MATLAB provides limited support for operations on uint8 arrays. imread reads images into MATLAB as uint8 arrays, and the image and imagesc functions can display uint8 images. MATLAB also supports these operations on uint8 arrays: • Displaying data values • Indexing into arrays using standard MATLAB subscripting • Reshaping, reordering, and concatenating arrays, using functions such as reshape, cat, and permute www.MatlabSite.com 1-13 1 Introduction • Saving to and loading from MAT-files • The all and any functions • Logical operators and indexing • Relational operators In addition, MATLAB supports the find function for uint8 arrays, but the returned array is of class double. Aside from the ones listed above, MATLAB does not support most operations for uint8 arrays. In particular, mathematical operations are not supported. If you attempt to perform an unsupported operation on a uint8 array, you receive an error. For example: BW3 = BW1 + BW2 ??? Function '+' not defined for variables of class 'uint8'. Most of the functions in the Image Processing Toolbox accept uint8 input. See the reference entries in Chapter 11 for information about the individual functions. Converting Betw een Data Types If you want to perform operations that are not supported for uint8 arrays, convert the data to double precision using the double function. For example: BW3 = double(BW1) + double(BW2); Keep in mind that converting between data types changes the way MATLAB and the toolbox interpret the image data. If you want the resulting array to be interpreted properly as image data, you need to rescale or offset the data when you convert it. The toolbox has two functions, im2double and im2uint8, for converting images from one data type to another. These functions automatically handle the rescaling and offsetting of the original data. For example, this command converts a double-precision RGB image with data in the range [0,1] to a uint8 RGB image with data in the range [0,255]: RGB2 = im2uint8(RGB1); To convert an indexed image, the input to the conversion function is the image matrix, not the colormap. Colormaps are always of class double in MATLAB. 1-14 www.MatlabSite.com W orking w ith Image Data Also, an indexed image cannot be converted from double to uint8 if the image matrix contains values greater than 256. For more information about im2double and im2uint8, see the reference entries for those functions in Chapter 11. Turning the Logical Flag on or off As discussed on page 1-7, a uint8 binary image must have its logical flag on. If you use im2uint8 to convert a binary image of type double to uint8, this flag is turned on automatically. If you do the conversion manually, however, you must use the logical function to turn on the logical flag. For example: B = logical(uint8(round(A))); To turn the logical flag off, you can use the unary plus operator. For example, if A is a uint8 logical array: B = +A; www.MatlabSite.com 1-15 1 Introduction Coordina te Systems Locations in an image can be expressed in various coordinate systems, depending on context. This section discusses the two main coordinate systems used in the Image Processing Toolbox, and the relationship between them. These systems are: • The pixel coordinate system • The spatial coordinate system Pix el Coordina tes Generally, the most convenient method for expressing locations in an image is to use pixel coordinates. In this coordinate system, the image is treated as a grid of discrete elements, ordered from top to bottom and left to right. For example: 1 2 3 c 1 2 3 r For pixel coordinates, the first component r (the row) increases downward, while the second component c (the column) increases to the right. Pixel coordinates are integer values and range between 1 and the length of the row or column. There is a one-to-one correspondence between pixel coordinates and the coordinates MATLAB uses for matrix subscripting. This correspondence makes the relationship between an image’s data matrix and the way the image displays easy to understand. For example, the data for the pixel in the fifth row, second column is stored in the matrix element (5,2). 1-16 www.MatlabSite.com C oordinate Systems Spa tia l Coordina tes In the pixel coordinate system, a pixel is treated as a discrete unit, uniquely identified by a single coordinate pair, such as (5,2). From this perspective, a location such as (5.3,2.2) is not meaningful. At times, however, it is useful to think of a pixel as a square patch, having area. From this perspective, a location such as (5.3,2.2) is meaningful, and is distinct from (5,2). In this spatial coordinate system, locations in an image are positions on a plane, and they are described in terms of x and y. This figure illustrates the spatial coordinate system used for images. Notice that y increases downward: 0.5 0.5 1 1.5 2 2.5 3 3.5 x 1 1.5 2 2.5 3 3.5 y This spatial coordinate system corresponds quite closely to the pixel coordinate system in many ways. For example, the spatial coordinates of the center point of any pixel are identical to the pixel coordinates for that pixel. There are some important differences, however. In pixel coordinates, the upper-left corner of an image is (1,1), while in spatial coordinates, this location by default is (0.5,0.5). This difference is due to the pixel coordinate system being discrete, while the spatial coordinate system is continuous. Also, the upper-left corner is always (1,1) in pixel coordinates, but you can specify a nondefault origin for the spatial coordinate system. See “Using a Nondefault Spatial Coordinate System” on page 1-18 for more information. Another potentially confusing difference is largely a matter of convention: the order of the horizontal and vertical components is reversed in the notation for these two systems. Pixel coordinates are expressed as (r,c), while spatial coordinates are expressed as (x,y). In the reference entries in Chapter 11, when www.MatlabSite.com 1-17 1 Introduction the syntax for a function uses r and c, it refers to the pixel coordinate system. When the syntax uses x and y, it refers to the spatial coordinate system. Using a N ondefault Spatial Coordinate System By default, the spatial coordinates of an image correspond with the pixel coordinates. For example, the center point of the pixel in row 5, column 3 has spatial coordinates x=3, y=5. (Remember, the order of the coordinates is reversed.) This correspondence simplifies many of the toolbox functions considerably. Several functions primarily work with spatial coordinates rather than pixel coordinates, but as long as you are using the default spatial coordinate system, you can specify locations in pixel coordinates. In some situations, however, you may want to use a nondefault spatial coordinate system. For example, you could specify that the upper-left corner of an image is the point (19.0,7.5), rather than (0.5,0.5). If you call a function that returns coordinates for this image, the coordinates returned will be values in this nondefault spatial coordinate system. To establish a nondefault spatial coordinate system, you can specify the XData and YData image properties when you display the image. These properties are two-element vectors that control the range of coordinates spanned by the image. By default, for an image A, XData is [1 size(A,2)], and YData is [1 size(A,1)]. For example, if A is a 100 row by 200 column image, the default XData is [1 200], and the default YData is [1 100]. The values in these vectors are actually the coordinates for the center points of the first and last pixels (not the pixel edges), so the actual coordinate range spanned is slightly larger; for instance, if XData is [1 200], the x-axis range spanned by the image is [0.5 200.5]. 1-18 www.MatlabSite.com C oordinate Systems These commands display an image using nondefault XData and YData: A = magic(5); x = [19.5 23.5]; y = [8.0 12.0]; image(A,'XData',x,'YData',y), axis image, colormap(jet(25)) See the reference entries in Chapter 11 for information about functions in the Image Processing Toolbox that can establish nondefault spatial coordinate systems. www.MatlabSite.com 1-19 1 Introduction 1-20 www.MatlabSite.com 2 Displaying Images Overview . . . . . . . . . . . . . . . . . . . . . 2-2 Standard Display Techniques . . . . . . . Displaying Images with imshow . . . . . . . Displaying Indexed Images . . . . . . . . . Displaying Intensity Images . . . . . . . . . Displaying Binary Images . . . . . . . . . . Displaying RGB Images . . . . . . . . . . . Displaying Images in Graphics Files . . . . . . Displaying Nonimage Data as an Intensity Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-3 2-3 2-5 2-5 2-6 2-7 2-8 2-8 Special Display Techniques . . . Adding a Colorbar . . . . . . . . Displaying Multiframe Image Arrays Displaying Multiple Images . . . . Zooming in on a Region of an Image . Texture Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-10 2-10 2-11 2-14 2-16 2-17 Printing Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-19 www.MatlabSite.com 2 Displaying Images O verview The Image Processing Toolbox supports a number of image display techniques. For example, the function imshow displays any image type with a single function call. Other functions handle more specialized display needs. This chapter describes basic display techniques and discusses topics such as multiple image display and texture mapping. In addition, it includes information about printing images. 2-2 www.MatlabSite.com Standard Display Techniques Sta nda rd Displa y Techniques In MATLAB, the primary way to display images is by using the image function. This function creates a Handle Graphics® image object, and it includes syntax for setting the various properties of the object. MATLAB also includes the imagesc function, which is similar to image but which automatically scales the input data. The Image Processing Toolbox includes an additional display routine, imshow. Like image and imagesc, this function creates a Handle Graphics image object. However, imshow also automatically sets various Handle Graphics properties and attributes of the image to optimize the display. This section discusses displaying images using imshow. In general, imshow is preferable to image and imagesc for image processing applications. For information about image and imagesc, see the online MATLAB Function Reference. These functions are also described in more detail in the Using MATLAB Graphics manual. Displa y ing Im a ges w ith im show When you display an image using the imshow function, MATLAB sets figure, axes, and image properties that control the way image data is interpreted. These properties include the image CData and CDataMapping property, the axes CLim property, and the figure Colormap property. In addition, depending on the arguments you specify and the settings of toolbox preferences, imshow may also: • Set other figure and axes properties to tailor the display. For example, axes and tick marks may not display. • Include or omit a border around the image. • Call the truesize function to display the image without interpolation. This section describes how you can control the behavior of imshow. The sections that follow it discuss how imshow displays specific image types. Preferences The Image Processing Toolbox includes a function, iptsetpref, that you can use to set preferences that affect the behavior of imshow: www.MatlabSite.com 2-3 2 Displaying Images • The ImshowBorder preference controls whether imshow leaves a border between the axes containing the image and the edges of the figure. • The ImshowAxesVisible preference controls whether imshow displays images with the axes box and tick labels. • The ImshowTruesize preference controls whether imshow calls the truesize function. To determine the current value of a preference, use the iptgetpref function. For more information about these and other toolbox preferences, and the values they accept, see the reference entries for iptsetpref and iptgetpref in Chapter 11. The truesize Function The truesize function assigns a single screen pixel to each image pixel. This is generally the preferred way to display an image. If you display an image without calling truesize, the image displays at the default axis size; MATLAB must use interpolation to determine the values for display pixels that do not correspond to elements in the image matrix. (See Chapter 3 for a discussion of interpolation.) In most situations, the imshow command calls truesize automatically. If you do not want imshow to call truesize (for example, if you display an image with a small number of pixels), you can set the ImshowTruesize preference to 'manual'. For example: iptsetpref('ImshowTruesize','manual') You can also override the setting of the ImshowTruesize preference for a single imshow command. For example, this command prevents imshow from calling truesize, regardless of the preference setting: imshow(X,map,'notruesize') For more information about the imshow and truesize functions, see the reference entries for those functions in Chapter 11. 2-4 www.MatlabSite.com Standard Display Techniques Displa y ing Index ed Im a ges To display an indexed image with imshow, you specify both the image matrix and the colormap: imshow(X,map) For each pixel in X, imshow displays the color stored in the corresponding row of map. imshow sets the Handle Graphics properties that control how colors display: • The image CData is set to the data in X. • The image CDataMapping property is set to direct. • The axes CLim property does not apply, because CDataMapping is set to direct. • The figure Colormap property is set to the data in map. The relationship between the values in the image matrix and the colormap depends on whether the image matrix is of class double or uint8. If the image matrix is of class double, the value 1 points to the first row in the colormap, the value 2 points to the second row, and so on. If the image matrix is of class uint8, there is an offset; the value 0 points to the first row in the colormap, the value 1 points to the second row, and so on. The offset is handled automatically by the image object, and is not controlled through a Handle Graphics property. Displa y ing Intensity Im a ges To display a grayscale intensity image, the syntax is: imshow(I) imshow displays the image by scaling the intensity values to serve as indices into a grayscale colormap. This is done as follows: • The image CData is set to the data in the image matrix. • The image CDataMapping property is set to scaled. • The axes CLim property is set to [0 1] if the image matrix is of class double, or [0 255] if the image matrix is of class uint8. • The figure Colormap property is set to a grayscale colormap whose values range from black to white. www.MatlabSite.com 2-5 2 Displaying Images By default, the number of levels of gray in the colormap is 256 on systems with 24-bit color, and 64 on other systems. You can explicitly specify the number of gray levels with imshow. For example, you can display an image I with 32 gray levels using: imshow(I,32) Note that this is not equivalent to: imshow(I,gray(32)) If you use the latter syntax, imshow interprets the values in I as indices into a colormap and therefore does not set CDataMapping to scaled. Because MATLAB scales intensity images to fill the colormap range, you can use any size colormap. Larger colormaps allow you to see more detail, but they also use up more color slots. For information about the number of color slots available on your system, see page 2-14. Displa y ing Bina r y Im a ges To display a binary image, the syntax is: imshow(BW) If the image matrix is of class double, imshow treats a binary image as an intensity image, setting CDataMapping to scaled, CLim to [0 1], and Colormap to a grayscale colormap. This means that the 0 values in the image matrix display as black, and the 1 values display as white. If the image matrix is of class uint8, the behavior depends on whether or not the matrix has its logical flag on: • If the logical flag is off, imshow treats the matrix as a uint8 intensity image, setting CLim to [0 255]. • If the logical flag is on, imshow sets CLim to [0 1]. In general, this means that a uint8 binary image must have its logical flag on to display properly. Therefore, all of the toolbox routines that return binary images return them as uint8 logical matrices. imread reads in an image as a uint8 logical array of 0’s and 1’s only if the image is stored in the file with 1 bit per pixel. If a binary image is stored in a file with 2-6 www.MatlabSite.com Standard Display Techniques 4 or 8 bits per pixel, imread interprets this as a grayscale intensity image; the resulting array has values of 0 and 255, and the logical flag is off. You can convert this image to a uint8 logical array using the NOT EQUAL (~=) operator in MATLAB: BW2 = BW ~= 0; Changing the Display Colors You may prefer to invert binary images when you display them, so that 0 values display as white and 1 values display as black. To do this, use the NOT (~) operator in MATLAB. For example: imshow(~BW) You can also display a binary image using a colormap. If the image is of class uint8, the 0 values display as the first color in the colormap, and the 1 values display as the second color. For example, the following command displays 0 as red and 1 as blue: imshow(BW,[1 0 0; 0 0 1]) If the image is of class double, you need to add 1 to each value in the image matrix, because there is no offset in the colormap indexing: imshow(BW + 1,[1 0 0; 0 0 1]) Displa y ing RGB Im a ges RGB images, also called truecolor images, represent color values directly, rather than through a colormap. To display an RGB image, the syntax is: imshow(RGB) RGB is m-by-n-by-3 array. For each pixel (r,c) in RGB, imshow displays the color represented by the triplet (r,c,1:3). As with other image types, MATLAB sets the image CData to the data in the image array. However, in this case CData is a three-dimensional array, rather than a two-dimensional matrix. If CData is three-dimensional, MATLAB interprets the array as truecolor data, and ignores the values of the CDataMapping, CLim, and Colormap properties. www.MatlabSite.com 2-7 2 Displaying Images An RGB image array can be of class double, in which case it contains values in the range [0,1], or of class uint8, in which case the data range is [0,255]. Systems that use 24 bits per screen pixel can display truecolor images directly, because they allocate 8 bits (256 levels) each to the red, green, and blue color planes. On systems with fewer colors, MATLAB displays the image using a combination of color approximation and dithering. See Chapter 10 for more information. Displa y ing Im a ges in Gra phics Files Generally, when you display an image, the image data is stored as one or more variables in the MATLAB workspace. However, if you have an image stored in a graphics file that imread can read, you can display a file directly using this syntax: imshow filename The file must be in the current directory or on the MATLAB path. For example, to display a file named flowers.tif: imshow flowers.tif This syntax is very useful for scanning through images stored in graphics files. Note, however, that when you use this syntax, the image data is not stored in the MATLAB workspace; if you want to process the image, you need to use imread to read in the data, or else use getimage to get the image data from the Handle Graphics image object. Also, if you use this syntax with a file containing more than one image, imshow displays only the first image. Displa y ing N onim a ge Da ta a s a n Intensity Im a ge In some cases, you may have data you want to display as an intensity image, even though the data is outside the appropriate range (i.e., [0,1] for double arrays or [0,255] for uint8 arrays). For example, if you filter an intensity image, the output data may not fall in the range of the original data, but you may still want to display the results as an image. To display this data as an image, you can specify the data range directly, using: imshow(I,[low high]) 2-8 www.MatlabSite.com Standard Display Techniques When you use this syntax, imshow sets the axes CLim property to [low high]. CDataMapping is always scaled for intensity images, so the image data is scaled so that low corresponds to the first row of the grayscale colormap and high corresponds to the last row. If you use an empty matrix ([]) for the data range, imshow scales the data automatically, setting low and high to the minimum and maximum values in the array. For example: I = imread('testpat1.tif'); J = filter2([1 2;–1 –2],I); imshow(J,[]) www.MatlabSite.com 2-9 2 Displaying Images Specia l Displa y Techniques In addition to imshow, the toolbox includes functions that perform specialized display operations, or exercise more direct control over the display format. These functions, together with the MATLAB graphics functions, provide a range of image display options. This section includes the following topics: • Adding a colorbar to an image display • Displaying multiframe image arrays • Displaying multiple images • Zooming in on a region of an image • Texture mapping an image onto a surface Adding a Colorba r You can use the colorbar function to add a colorbar to an axes object. If you add a colorbar to an axes object that contains an image object, the colorbar indicates the data values that the different colors in the image correspond to. Seeing the correspondence between data values and the colors displayed is especially useful if you are displaying nonimage data as an image, as described 2-10 www.MatlabSite.com Special Display Techniques on page 2-8. In the example below, you filter a grayscale image of class uint8, and the resulting data is no longer in the range [0,255]: I = imread('saturn.tif'); h = [1 2 1; 0 0 0; –1 –2 –1]; I2 = filter2(h,I); imshow(I2,[]), colorbar Displa y ing M ultifra m e Im a ge Arra y s To view the images in a multiframe array, you can use any of these options: • Display the frames individually, using the imshow function • Display all of the frames at once, using the montage function • Convert the array to a MATLAB movie, using the immovie function Displaying Frames Individually To view an individual frame, call imshow and specify the frame using standard MATLAB indexing notation. For example, to view the seventh frame in the intensity array I: imshow(I(:,:,:,7)) www.MatlabSite.com 2-11 2 Displaying Images Displaying All Frames at O nce To view all of the frames in a multiframe array at one time, use the montage function. montage divides a figure into multiple display regions and displays each image in a separate region. The syntax for montage is similar to the imshow syntax. To display a multiframe intensity image, the syntax is: montage(I) To display a multiframe indexed image, the syntax is: montage(X,map) Note that all of the frames in a multiframe indexed array must use the same colormap. The commands below display a multiframe image array. Notice that montage displays images in a row-wise manner. The first frame appears in the first 2-12 www.MatlabSite.com Special Display Techniques position of the first row, the next frame in the second position of the first row, and so on. montage arranges the frames so that they roughly form a square. load mri montage(D,map) Converting the Array to a M ovie To create a MATLAB movie from a multiframe image array, use the immovie function. This function works only with indexed images; if you want to convert another type of image array to a movie, you must first convert it to an indexed image, using one of the conversion functions described in Chapter 1. This call creates a movie from the multiframe indexed image shown above: mov = immovie(D,map); www.MatlabSite.com 2-13 2 Displaying Images You can play the movie in MATLAB using the movie function: movie(mov) Note that when you play the movie, you need to supply the colormap used by the original image array. For example: mov = immovie(D,map); colormap(map), movie(mov) Displa y ing M ultiple Im a ges MATLAB does not place any restrictions on the number of images you can display simultaneously. However, there are usually system limitations that are dependent on the computer hardware you are using. The main limitation is the number of colors your system can display. This number depends primarily on the number of bits that are used to store the color information for each pixel. Most systems use either 8, 16, or 24 bits per pixel. If you are using a system with 16 or 24 bits per pixel, you are unlikely to run into any problems, regardless of the number of images you display. However, if your system uses 8 bits per pixel, it can display a maximum of 256 different colors, so you can quickly run out of color slots if you display multiple images. (Actually, the total number of colors you can display is slightly fewer than 256, because some color slots are reserved for Handle Graphics objects. The operating system usually reserves a few colors as well.) To determine the number of bits per pixel on your system, enter this command: get(0,'ScreenDepth') See Chapter 10 for more information about working with different color depths. This section discusses methods you can use to display multiple images at the same time, and includes information about working around system limitations. This section discusses: • Displaying each image in a separate figure • Displaying multiple images in the same figure Displaying Each Image in a Separate Figure The simplest way to display multiple images is to display them in different figure windows. imshow always displays in the current figure, so if you display 2-14 www.MatlabSite.com Special Display Techniques two images in succession, the second image replaces the first image. Therefore, if you want to display two or more images in separate windows, you must explicitly create a new empty figure before calling imshow for the second and subsequent images. For example: imshow(I) figure, imshow(I2) When you use this approach, the figures you create are empty initially, so the images are displayed using truesize. If you have an 8-bit display, you must make sure that the total number of colormap entries does not exceed 256. For example, if you try to display three images, each having a different colormap with 128 entries, at least one of the images will display with the wrong colors. (If all three images use the same colormap, there will not be a problem, because only 128 color slots are used.) Note that intensity images are displayed using colormaps, so the color slots used by these image count toward the 256-color total. One way to avoid these display problems is to manipulate the colormaps to use fewer colors. There are various ways to do this, such as using the imapprox function. See Chapter 10 for information about reducing the number of colors used by an image. Another solution is to convert images to RGB (truecolor) format for display. The colors used by truecolor images do not count toward the 256-color total, because MATLAB automatically uses dithering and color approximation to display these images. For example, to display an indexed image as an RGB image: imshow(ind2rgb(X,map)) To display an intensity image as an RGB image: imshow(cat(3,I,I,I)) Displaying M ultiple Images in the Same Figure You can display multiple images in a single figure window by using imshow in conjunction with the subplot function. This function divides a figure into multiple display regions. The syntax of subplot is: subplot(m,n,p) www.MatlabSite.com 2-15 2 Displaying Images This call divides the figure into an m-by-n matrix of display regions and makes the p-th display region active. For example, you can display two images side by side using: subplot(1,2,1), imshow(X1,map) subplot(1,2,2), imshow(X2,map) To display multiple images in the same figure with imshow, the images must use the same colormap, as in the example above. If the colormaps differ, the images will not display properly. This is because there is only one colormap for a given figure, and this figure colormap is replaced each time you use imshow. For example, if you display a second image in the figure, that image’s colormap will become the figure colormap, and the first image will use it as well, even if it is not appropriate. To avoid this problem, the toolbox provides a function, subimage, that you can use instead of imshow to display images in subplots. For example, to display two images in the same figure: subplot(1,2,1), subimage(X1,map1) subplot(1,2,2), subimage(X2,map2) Both images display correctly, even if they have different colormaps. Generally, subimage converts the images to RGB format for display purposes. Z oom ing in on a Region of a n Im a ge The zoom command enables you to examine the details of an image by using the mouse to zoom in or out. When you zoom in, the figure remains the same size, but only a portion of the image is displayed, at a higher magnification. (zoom works by changing the axis limits; it does not change the image data in the figure.) To turn zoom mode on, type: zoom on There are two ways to zoom in on an image: 2-16 www.MatlabSite.com Special Display Techniques • Click on a spot in the image by placing the cursor on the spot and the pressing the left mouse button. The view changes automatically. The center of the new view is the spot where you clicked. • Select a region by clicking on the image, holding down the mouse button, and dragging the mouse. A dotted rectangle appears. When you release the mouse button, the region enclosed by the rectangle is displayed. To zoom out, click on the image with the right mouse button. (If you have a single-button mouse, hold down the Shift key and click.) To zoom out completely and restore the original view, enter: zoom out To turn zoom mode off: zoom off Tex ture M a pping When you use the imshow command, MATLAB displays the image in a two-dimensional view. However, it is also possible to display an image on a parametric surface, such as a sphere, or below a surface plot. The warp function creates these displays by texture mapping the image. Texture mapping is a process that maps an image onto a surface grid using bilinear interpolation. www.MatlabSite.com 2-17 2 Displaying Images This example texture maps an image of a test pattern onto a cylinder: [x,y,z] = cylinder; I = imread('testpat1.tif'); warp(x,y,z,I); For more information about texture mapping, see the reference entry for the warp function in Chapter 11. 2-18 www.MatlabSite.com Printing Images Printing Ima ges If you want to output a MATLAB image to use it in another application (such as a word-processing program or graphics editor), use imwrite to create a file in the appropriate format. If you want to print the contents of a MATLAB figure (including nonimage elements such as labels), use the MATLAB print command, or choose the Print option from the File menu of the figure window. If you produce output this way, the results reflect the settings of various Handle Graphics properties. In some cases, you may need to change the settings of certain properties to get the results you want. Here are some tips that may be helpful when you print images: • Image colors print as shown on the screen. This means that images do not obey the InvertHardcopy figure property. • To ensure that printed images have the proper size and aspect ratio, you should set the figure’s PaperPositionMode property to auto. When PaperPositionMode is set to auto, the width and height of the printed figure are determined by the figure’s dimensions on the screen. By default, the value of PaperPositionMode is manual. If you want the default value of PaperPositionMode to be auto, enter this line in your startup.m file: set(0,'DefaultFigurePaperPositionMode','auto') For more information about printing and about Handle Graphics, see the Using MATLAB Graphics manual. www.MatlabSite.com 2-19 2 Displaying Images 2-20 www.MatlabSite.com 3 Geometric Operations Overview . . . . . . . . . . . . . . . . . . . . . 3-2 Interpolation . . . . . . . . . . . . . . . . . . . 3-3 Image Types . . . . . . . . . . . . . . . . . . . . 3-4 Image Resizing . . . . . . . . . . . . . . . . . . . 3-5 Image Rotation . . . . . . . . . . . . . . . . . . . 3-6 Image Cropping . . . . . . . . . . . . . . . . . . 3-7 www.MatlabSite.com 3 G eometric O perations O verview This chapter describes some basic image processing tools, the geometric functions. These functions modify the geometry of an image by resizing, rotating, or cropping the image. These functions support all image types. The chapter begins with a discussion of interpolation, an operation common to most of the geometric functions. It then discusses each of the geometric functions separately, and shows how to apply them to sample images. 3-2 www.MatlabSite.com Interpolation Interpola tion The imresize and imrotate geometric functions use two-dimensional interpolation as part of the operations they perform. (The improfile image analysis function also uses interpolation. See Chapter 7 for information about this function.) Interpolation is the process by which the software resamples the image data to determine values between defined pixels. For example, if you resize an image so it contains more pixels than it did originally, the software obtains values for the additional pixels through interpolation. The Image Processing Toolbox provides three interpolation methods: • Nearest neighbor interpolation • Bilinear interpolation • Bicubic interpolation The interpolation methods all work in a fundamentally similar way. In each case, to determine the value for an interpolated pixel, you find the point in the input image that the output pixel corresponds to. You then assign a value to the output pixel by computing a weighted average of some set of pixels in the vicinity of the point. The weightings are based on the distance each pixel is from the point. The methods differ in the set of pixels that are considered: • For nearest neighbor interpolation, the output pixel is assigned the value of the pixel that the point falls within. No other pixels are considered. • For bilinear interpolation, the output pixel value is a weighted average of pixels in the nearest 2-by-2 neighborhood. • For bicubic interpolation, the output pixel value is a weighted average of pixels in the nearest 4-by-4 neighborhood. The number of pixels considered affects the complexity of the computation. Therefore the bilinear method takes longer than nearest neighbor interpolation, and the bicubic method takes longer than bilinear. However, the greater the number of pixels considered, the more accurate the effect is, so there is a tradeoff between processing time and quality. www.MatlabSite.com 3-3 3 G eometric O perations Im a ge Ty pes The functions that use interpolation take an argument that specifies the interpolation method. For these functions, the default method is nearest neighbor interpolation. This method produces acceptable results for all image types, and is the only method that is appropriate for indexed images. For intensity and RGB images, however, you should generally specify bilinear or bicubic interpolation, because these methods produce better results than nearest neighbor interpolation. For RGB images, interpolation is performed on the red, green, and blue image planes individually. For binary images, interpolation has effects that you should be aware of. If you use bilinear or bicubic interpolation, the computed values for the pixels in the output image will not all be 0 or 1. The effect on the resulting output image depends on the class of the input image: • If the class of the input image is double, the output image is a grayscale image of class double. The output image is not binary, because it includes values other than 0 and 1. • If the class of the input image is uint8, the output image is a binary image of class uint8. The interpolated pixel values are rounded off to 0 and 1 so the output image can be of class uint8. If you use nearest neighbor interpolation, the result is always binary, because the values of the interpolated pixels are taken directly from pixels in the input image. 3-4 www.MatlabSite.com Image Resizing Ima ge Resizing The toolbox function imresize changes the size of an image using a specified interpolation method. If you do not specify an interpolation method, the function uses nearest neighbor interpolation. You can use imresize to resize an image by a specific magnification factor. To enlarge an image, specify a factor greater than 1. For example, this command doubles the number of pixels in X in each direction: Y = imresize(X,2) To reduce an image, specify a number between 0 and 1 as the magnification factor. You can also specify the actual size of the output image. This command creates an output image of size 100-by-150: Y = imresize(X,[100 150]) If the specified size does not produce the same aspect ratio as the input image has, the output image will be distorted. If you reduce the image size and use bilinear or bicubic interpolation, imresize applies a lowpass filter to the image before interpolation. This reduces the effect of Moiré patterns, ripple patterns that result from aliasing during resampling. Note, however, that even with lowpass filtering, the resizing operation can introduce artifacts, because information is always lost when you reduce the size of an image. imresize does not apply a lowpass filter if nearest neighbor interpolation is used, unless you explicitly specify the filter. This interpolation method is primarily used for indexed images, and lowpass filtering is not appropriate for these images. For information about specifying a different filter, see the reference entry for imresize in Chapter 11. www.MatlabSite.com 3-5 3 G eometric O perations Ima ge Rota tion The imrotate function rotates an image, using a specified interpolation method and rotation angle. If you do not specify an interpolation method, the function uses nearest neighbor interpolation. You specify the rotation angle in degrees. If you specify a positive value, imrotate rotates the image counterclockwise; if you specify a negative value, imrotate rotates the image clockwise. For example, these commands rotate an image 35˚ counterclockwise: I = imread('ic.tif'); J = imrotate(I,35,'bilinear'); imshow(I) figure, imshow(J) Notice that the resulting image is larger than the input image, in order to include the entire original image. imrotate adds 0’s to the image to fill the area outside the original image. imrotate has an option for cropping the output image to the same size as the input image. See the reference entry for imrotate in Chapter 11 for more information. 3-6 www.MatlabSite.com Image C ropping Ima ge Cropping The function imcrop extracts a rectangular portion of an image. You can specify the crop rectangle through input arguments, or select it with a mouse. If you call imcrop without specifying the crop rectangle, the cursor changes to a cross hair when it is over the image. Click on one corner of the region you want to select, and while holding down the mouse button, drag across the image. imcrop draws a rectangle around the area you are selecting. When you release the mouse button, imcrop creates a new image from the selected region. In this example, you display an image and call imcrop. The rectangle you select is shown in red. imshow ic.tif I = imcrop; Now display the cropped image: imshow(I) If you do not provide any output arguments, imcrop displays the image in a new figure. www.MatlabSite.com 3-7 3 G eometric O perations 3-8 www.MatlabSite.com 4 Neighborhood and Block Operations Overview . . . . . . . . . . . . . . . . . . . . . 4-2 Types of Block Processing Operations . . . . . . . . . . 4-2 Sliding Neighborhood Operations . . . . . . . . . . 4-4 Padding of Borders . . . . . . . . . . . . . . . . . . 4-5 Linear and Nonlinear Filtering . . . . . . . . . . . . . 4-5 Distinct Block Operations . . . . . . . . . . . . . . 4-8 Overlap . . . . . . . . . . . . . . . . . . . . . . 4-9 Column Processing . . . . . . . . . . . . . . . . . 4-11 Sliding Neighborhoods . . . . . . . . . . . . . . . . 4-11 Distinct Blocks . . . . . . . . . . . . . . . . . . . 4-12 www.MatlabSite.com 4 N eighborhood and Block O perations O verview Certain image processing operations involve processing an image in sections called blocks, rather than processing the entire image at once. For example, many linear filtering operations and binary image operations work this way. The Image Processing Toolbox provides several functions for specific operations that work with blocks; for example, the dilate function for binary image dilation. In addition, the toolbox provides more generic functions for processing an image in blocks. This chapter discusses these generic block processing functions. To use one of the functions described in this chapter, you supply information about the size of the blocks, and specify a separate function to use to process the blocks. The block processing function does the work of breaking the input image into blocks, calling the specified function for each block, and reassembling the results into an output image. Ty pes of Block Processing O pera tions Using these functions, you can perform various block processing operations, including sliding neighborhood operations and distinct block operations: • In a sliding neighborhood operation, the input image is processed in a pixelwise fashion. That is, for each pixel in the input image, some operation is performed to determine the value of the corresponding pixel in the output image. The operation is based on the values of a block of neighboring pixels. • In a distinct block operation, the input image is processed a block at a time. That is, the image is divided into rectangular blocks, and some operation is performed on each block individually to determine the values of the pixels in the corresponding block of the output image. In addition, the toolbox provides functions for column processing operations. These operations are not actually distinct from block operations; instead, they are a way of speeding up block operations by rearranging blocks into matrix columns. Note that even if you do not use the block processing functions described in this chapter, the information here may be useful to you, as it includes concepts fundamental to many areas of image processing. In particular, the discussion of sliding neighborhood operations is applicable to linear filtering and binary 4-2 www.MatlabSite.com O verview morphological operations. See Chapter 5 and Chapter 8 for information about these applications. www.MatlabSite.com 4-3 4 N eighborhood and Block O perations Sliding N eighborhood O pera tions A sliding neighborhood operation is an operation that is performed a pixel at a time, with the value of any given pixel in the output image being determined by applying some algorithm to the values of the corresponding input pixel’s neighborhood. A pixel’s neighborhood is some set of pixels, defined by their locations relative to that pixel, which is called the center pixel. The neighborhood is a rectangular block, and as you move from one element to the next in an image matrix, the neighborhood block slides in the same direction. The figure below shows the neighborhood blocks for some of the elements in a 6-by-5 matrix with 2-by-3 sliding blocks. The center pixel for each neighborhood is marked with a dot. The center pixel is the actual pixel in the input image being processed by the operation. If the neighborhood has an odd number of rows and columns, the center pixel is actually in the center of the neighborhood. If one of the dimensions has even length, the center pixel is just to the left of center or just above center. For example, in a 2-by-2 neighborhood, the center pixel is the upper left one. For any m-by-n neighborhood, the center pixel is: floor(([m n]+1)/2) In the 2-by-3 block shown in the figure above, the center pixel is (1,2), the pixel in the second column of the top row of the neighborhood. A sliding neighborhood operation proceeds like this: 4-4 www.MatlabSite.com Sliding N eighborhood O perations 1 Select a single pixel. 2 Determine the pixel’s neighborhood. 3 Apply a function to the values of the pixels in the neighborhood. This function must return a scalar. 4 Find the pixel in the output image whose position corresponds to that of the center pixel in the input image. Set this output pixel to the value returned by the function. 5 Repeat steps 1 through 4 for each pixel in the input image. For example, suppose the figure above represents an averaging operation. The function might sum the values of the six neighborhood pixels and then divide by 6. The result is the value of the output pixel. Pa dding of Borders As the figure on page 4-4 shows, some of the pixels in a neighborhood may be missing, especially if the center pixel is on the border of the image. Notice that in the figure, the upper left and bottom right neighborhoods include “pixels” that are not part of the image. To process these neighborhoods, sliding neighborhood operations pad the borders of the image, usually with 0’s. In other words, these functions process the border pixels by assuming that the image is surrounded by additional rows and columns of 0’s. These rows and columns do not become part of the output image and are used only as parts of the neighborhoods of the actual pixels in the image. Linea r a nd N onlinea r Filtering You can use sliding neighborhood operations to implement many kinds of filtering operations. One example of a sliding neighbor operation is convolution, which is used to implement linear filtering. MATLAB provides the conv2 and filter2 functions for performing convolution. See Chapter 5 for more information about these functions. In addition to convolution, there are many other filtering operations you can implement through sliding neighborhoods. Many of these operations are nonlinear in nature. For example, you can implement a sliding neighborhood www.MatlabSite.com 4-5 4 N eighborhood and Block O perations operation where the value of an output pixel is equal to the standard deviation of the values of the pixels in the input pixel’s neighborhood. You can use the nlfilter function to implement a variety of sliding neighborhood operations. nlfilter takes as input arguments an image, a neighborhood size, and a function that returns a scalar, and returns an image of the same size as the input image. The value of each pixel in the output image is computed by passing the corresponding input pixel’s neighborhood to the function. For example, this call computes each output pixel by taking the standard deviation of the values of the input pixel’s 3-by-3 neighborhood (that is, the pixel itself and its eight contiguous neighbors): I2 = nlfilter(I,[3 3],'std2'); You can write an M-file to implement a specific function, and then use this function with nlfilter. For example, this command processes the matrix I in 2-by-3 neighborhoods with a function called myfun.m: nlfilter(I,[2 3],'myfun'); You can also use an inline function; in this case, the function name appears in the nlfilter call without quotation marks. For example: f = inline('sqrt(min(x(:)))'); I2 = nlfilter(I,[2 2],f); 4-6 www.MatlabSite.com Sliding N eighborhood O perations The example below uses nlfilter to set each pixel to the maximum value in its 3-by-3 neighborhood: I = imread('tire.tif'); f = inline('max(x(:))'); I2 = nlfilter(I,[3 3],f); imshow(I) figure, imshow(I2) Many operations that nlfilter can implement run much faster if the computations are performed on matrix columns rather than rectangular neighborhoods. For information about this approach, see the discussion of the colfilt function on page 4-11. www.MatlabSite.com 4-7 4 N eighborhood and Block O perations Distinct Block O pera tions Distinct blocks are rectangular partitions that divide a matrix into m-by-n sections. Distinct blocks overlay the image matrix starting in the upper-left corner, with no overlap. If the blocks don’t fit exactly over the image, the toolbox adds zero padding so that they do. The figure below shows a 15-by-30 matrix divided into 4-by-8 blocks. The zero padding process adds 0’s to the bottom and right of the image matrix, as needed. After zero padding, the matrix is size 16-by-32. The function blkproc performs distinct block operations. blkproc extracts each distinct block from an image and passes it to a function you specify. blkproc assembles the returned blocks to create an output image. For example, the command below processes the matrix I in 4-by-6 blocks with the function myfun. I2 = blkproc(I,[4 6],'myfun'); You can specify the function as an inline function; in this case, the function name appears in the blkproc call without quotation marks. For example: f = inline('mean2(x)*ones(size(x))'); I2 = blkproc(I,[4 6],f); 4-8 www.MatlabSite.com Distinct Block O perations The example below uses blkproc to set every pixel in each 8-by-8 block of an image matrix to the average of the elements in that block. I = imread('tire.tif'); f = inline('uint8(round(mean2(x)*ones(size(x))))'); I2 = blkproc(I,[8 8],f); imshow(I) figure, imshow(I2); Notice that the inline function computes the mean of the block and then multiplies the result by a matrix of ones, so that the output block is the same size as the input block. As a result, the output image is the same size as the input image. blkproc does not require that the images be the same size; however, if this is the result you want, you must make sure that the function you specify returns blocks of the appropriate size. O verla p When you call blkproc, you can specify an overlap to the distinct blocks; that is, extra rows and columns of pixels outside the block whose values are taken into account when processing the block. When there is an overlap, blkproc passes the expanded block (including the overlap) to the specified function. www.MatlabSite.com 4-9 4 N eighborhood and Block O perations The figure below shows two 4-by-8 blocks with a 1-by-2 overlap. Notice that there is a 1-row overlap on each side of the block, and a 2-row overlap above and below the block. To specify the overlap, you provide an additional input argument to blkproc. To process the blocks in the figure above with the function myfun, the call is: B = blkproc(A,[4 8],[1 2],'myfun') Overlap often increases the amount of zero padding needed. For example, in the figure on page 4-8, the original 15-by-30 matrix becomes 16-by-32 with zero padding. If a 1-by-2 overlap is used, the padded matrix is 18-by-36. 4-10 www.MatlabSite.com C olumn Processing Column Processing The toolbox provides functions that you can use to process sliding neighborhoods or distinct blocks as columns. This approach is useful for operations that MATLAB performs columnwise; in many cases, column processing can reduce the execution time required to process an image. For example, suppose the operation you are performing involves computing the mean of each block. This computation is much faster if you first rearrange the blocks into columns, because you can compute the mean of every column with a single call to the mean function, rather than calling mean for each block individually. You can use the colfilt function to implement column processing. This function: 1 Reshapes each sliding or distinct block of an image matrix into a column in a temporary matrix 2 Passes the temporary matrix to a function you specify 3 Rearranges the resulting matrix back into the original shape Sliding N eighborhoods For a sliding neighborhood operation, colfilt creates a temporary matrix that has a separate column for each pixel in the original image. The column corresponding to a given pixel contains the values of that pixel’s neighborhood from the original image. The figure below illustrates this process. In this figure, a 6-by-5 image matrix is processed in 2-by-3 neighborhoods. colfilt creates one column for each pixel in the image, so there are a total of 30 columns in the temporary matrix. Each pixel’s column contains the value of the pixels in its neighborhood, so there are six rows. colfilt zero pads the input image as necessary. For www.MatlabSite.com 4-11 4 N eighborhood and Block O perations example, the neighborhood of the upper left pixel in the figure has two zero-valued neighbors, due to zero padding. The temporary matrix is passed to a function, which must return a single value for each column. (Many MATLAB functions work this way; for example, mean, median, std, sum, etc.) The resulting values are then assigned to the appropriate pixels in the output image. colfilt can produce the same results as nlfilter with faster execution time; however, it may use more memory. The example below sets each output pixel to the maximum value in the input pixel’s neighborhood, producing the same result as the nlfilter example on page 4-7. Notice that the function is max(x) rather than max(x(:)), because each neighborhood in the original image is a separate column in the temporary matrix. f = inline('max(x)'); I2 = colfilt(I,[3 3],'sliding',f); Distinct Block s For a distinct block operation, colfilt creates a temporary matrix by rearranging each block in the image into a column. colfilt pads the original image with 0’s, if necessary, before creating the temporary matrix. The figure below illustrates this process. In this figure, a 6-by-16 image matrix is processed in 4-by-6 blocks. colfilt first zero pads the image to make the size 4-12 www.MatlabSite.com C olumn Processing 8-by-18 (six 4-by-6 blocks), and then rearranges the blocks into 6 columns of 24 elements each. After rearranging the image into a temporary matrix, colfilt passes this matrix to the function. The function must return a matrix of the same size as the temporary matrix. If the block size is m-by-n, and the image is mm-by-nn, the size of the temporary matrix is (m*n)-by-(ceil(mm/m)*ceil(nn/n)). After the function processes the temporary matrix, the output is rearranged back into the shape of the original image matrix. This example sets all the pixels in each 8-by-8 block of an image to the mean pixel value for the block, producing the same result as the blkproc example on page 4-9: f = inline('ones(64,1)*mean(x)'); I2 = colfilt(I,[8 8],'distinct',f); www.MatlabSite.com 4-13 4 N eighborhood and Block O perations Notice that the inline function computes the mean of the block and then multiplies the result by a vector of ones, so that the output block is the same size as the input block. As a result, the output image is the same size as the input image. Restrictions You can use colfilt to implement many of the same distinct block operations that blkproc performs. However, colfilt has certain restrictions that blkproc does not: • The output image must be the same size as the input image. • The blocks cannot overlap. For situations that do not satisfy these constraints, use blkproc. 4-14 www.MatlabSite.com 5 Linear Filtering and Filter Design Overview . . . . . . . . . . . . . . . . . . . . . 5-2 Linear Filtering . . . . . . Convolution . . . . . . . . . Padding of Borders . . . . . . The filter2 Function . . . . . Separability . . . . . . . . . Higher-Dimensional Convolution Using Predefined Filter Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-3 5-3 5-5 5-7 5-8 5-9 5-9 Filter Design . . . . . . . . . . . . . . . FIR Filters . . . . . . . . . . . . . . . . Frequency Transformation Method . . . . . . Frequency Sampling Method . . . . . . . . . Windowing Method . . . . . . . . . . . . . Creating the Desired Frequency Response Matrix Computing the Frequency Response of a Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-13 5-13 5-14 5-15 5-16 5-18 5-19 www.MatlabSite.com 5 Linear Filtering and Filter Design O verview The Image Processing Toolbox provides a number of functions for designing and implementing two-dimensional linear filters for image data. This chapter describes these functions and how to use them effectively. The material in this chapter is divided into two parts: • The first part is an explanation of linear filtering and how it is implemented in the toolbox. This topic describes filtering in terms of the spatial domain, and is accessible to anyone doing image processing. • The second part is a discussion about designing two-dimensional Finite Infinite Response (FIR) filters. This section assumes you are familiar with working in the frequency domain. 5-2 www.MatlabSite.com Linear Filtering Linea r Filtering Filtering is a technique for modifying or enhancing an image. For example, you can filter an image to emphasize certain features or remove other features. Filtering is a neighborhood operation, in which the value of any given pixel in the output image is determined by applying some algorithm to the values of the pixels in the neighborhood of the corresponding input pixel. A pixel’s neighborhood is some set of pixels, defined by their locations relative to that pixel. (For a general discussion of neighborhood operations, see Chapter 4.) Linear filtering is filtering in which the value of an output pixel is a linear combination of the values of the pixels in the input pixel’s neighborhood. For example, an algorithm that computes a weighted average of the neighborhood pixels is one type of linear filtering operation. This section discusses linear filtering in MATLAB and the Image Processing Toolbox. It includes: • A description of how MATLAB performs linear filtering, using convolution • A discussion about using predefined filter types For information about how to design filters, see page 5-13. Convolution In MATLAB, linear filtering of images is implemented through two-dimensional convolution. In convolution, the value of an output pixel is computed by multiplying elements of two matrices and summing the results. One of these matrices represents the image itself, while the other matrix is the filter. For example, a filter might be: k = [4 –3 4 6 1 2] This filter representation is known as a convolution kernel. The MATLAB function conv2 implements image filtering by applying your convolution kernel to an image matrix. conv2 takes as arguments an input image and a filter, and returns an output image. For example, in this call, k is the convolution kernel, A is the input image, and B is the output image: B = conv2(A,k); www.MatlabSite.com 5-3 5 Linear Filtering and Filter Design conv2 produces the output image by performing these steps: 1 Rotate the convolution kernel 180 degrees to produce a computational molecule. 2 Determine the center pixel of the computational molecule. 3 Apply the computational molecule to each pixel in the input image. Each of these steps is explained below. Rotating the Convolution Kernel In two-dimensional convolution, the computations are performed using a computational molecule. This is simply the convolution kernel rotated 180 degrees, as in this call: h = rot90(k,2); h = 2 6 1 –3 4 4 Determining the Center Pixel To apply the computational molecule, you must first determine the center pixel. The center pixel is defined as floor((size(h)+1)/2). For example, in a 5-by-5 molecule, the center pixel is (3,3). The molecule h shown above is 2-by-3, so the center pixel is (1,2). Applying the Computational M olecule The value of any given pixel in B is determined by applying the computational molecule h to the corresponding pixel in A. You can visualize this by overlaying h on A, with the center pixel of h over the pixel of interest in A. You then multiply each element of h by the corresponding pixel in A, and sum the results. 5-4 www.MatlabSite.com Linear Filtering For example, to determine the value of the pixel (4,6) in B, overlay h on A, with the center pixel of h covering the pixel (4,6) in A. The center pixel is circled in this figure: A –2 4 –4 –4 –4 –2 4 –1 1 –4 –4 0 2 3 5 –2 –3 1 2 –4 2 0 –1 3 0 0 –1 2 1 1 3 2 4 5 2 6 3 1 h 4 3 –3 4 5 2 5 5 3 2 0 5 0 –1 –4 4 3 5 –4 –2 –4 –2 1 –1 1 –2 –1 2 –2 5 –4 Now, look at the six pixels covered by h. For each of these pixels, multiply the value of the pixel by the value in h. Sum the results, and place this sum in B(4,6): B(4,6) = 2*2 + 3*6 + 3*4 + 3*1 + 2*–3 + 0*4 = 31 Perform this procedure for each pixel in A to determine the value of each corresponding pixel in B. Pa dding of Borders When you apply a filter to pixels on the borders of an image, some of the elements of the computational molecule may not overlap actual image pixels. For example, if the molecule is 3-by-3 and you are computing the result for a pixel on the top row of the image, some of the elements of the molecule are outside the border of the image. www.MatlabSite.com 5-5 5 Linear Filtering and Filter Design This figure illustrates a 3-by-3 computational molecule being applied to the pixel (1,3) of a 5-by-5 matrix. The center pixel is indicate by a filled circle: In order to compute output values for the border pixels, conv2 pads the image matrix with zeroes. In other words, the output values are computed by assuming that the input image is surrounded by additional rows and columns of zeroes. In the figure shown above, the elements in the top row of the computational molecule are assumed to overlap zeroes. Depending on what you are trying to accomplish, you may want to discard output pixels whose values depend on zero padding. To indicate what portion of the convolution to return, conv2 takes a third input argument, called the shape parameter, whose value is one of these three strings: • 'valid' – returns only the pixels whose values can be computed without using zero padding of the input image. The resulting output image is smaller than the input image. In this example, the output image is 3-by-3. • 'same' – returns the set of pixels that can be computed by applying the filter to all pixels that are actually part of the input image. Border pixels are computed using zero padding, but the center pixel of the computational kernel is applied only to pixels in the image. This results in an output image that is the same size as the input image. • 'full' – returns the full convolution. This means conv2 returns all pixels for which any of the pixels in the computational molecule overlap pixels in the image, even when the center pixel is outside the input image. The resulting output image is larger than the input image. In this example, the output image is 7-by-7. conv2 returns the full convolution by default. 5-6 www.MatlabSite.com Linear Filtering The figure below illustrates applying a computational molecule to three different places in an image matrix: The com putational m olecule over- The com putational m olecule over- The com putational m olecule over- laps only pixels that are in laps pixels outside the original laps pixels at the edges only. The the original im age. The result is im age, but the center pixel overlaps center pixel is outside the im age. included in the output m atrix, a pixel in the im age. The result is The result is included in the output regardless of the shape param eter. included in the output m atrix if the m atrix if the shape param eter is shape param eter is same or full. full. If you use the full option, then the order of the first two input arguments is interchangeable, because full convolution is commutative. In other words, it does not matter which matrix is considered the convolution kernel, because the result is the same in either case. If you use the valid or same option, the operation is not commutative, so the convolution kernel must be the second argument. The filter2 Function In addition to the conv2 function, MATLAB also provides the filter2 function for two-dimensional linear filtering. filter2 can produce the same results as conv2, and differs primarily in that it takes a computational molecule as an input argument, rather than a convolution kernel. (filter2 operates by forming the convolution kernel from the computational molecule and then calling conv2.) The operation that filter2 performs is called correlation. www.MatlabSite.com 5-7 5 Linear Filtering and Filter Design If k is a convolution kernel, h is the corresponding computational molecule, and A is an image matrix, these calls produce identical results: B = conv2(A,k,'same'); and: B = filter2(h,A,'same'); The functions in the Image Processing Toolbox that produce filters (fspecial, fsample, etc.) all return computational molecules. You can use these filters directly with filter2, or you can rotate them 180 degrees and call conv2. Sepa ra bility Before calling conv2 to perform two-dimensional convolution, filter2 first checks whether the filter is separable into two one-dimensional filters (one column vector and one row vector). If the filter is separable, filter2 uses singular value decomposition to find the two vectors. filter2 then calls conv2 with this syntax: conv2(A,kcol,krow); kcol and krow are the column and row vectors that the two-dimensional convolution kernel k separates into (that is, k = kcol*krow). conv2 filters the columns with the column vector, and then, using the output of this operation, filters the rows using the row vector. The result is equivalent to two-dimensional convolution but is faster because it requires fewer computations. 5-8 www.MatlabSite.com Linear Filtering Determining Separability A filter is separable if its rank is 1. For example, this filter is separable: k = 1 2 4 2 4 8 3 6 12 rank(k) ans = 1 H igher-Dim ensiona l Convolution To perform two-dimensional convolution, you use conv2 or filter2. To perform higher-dimensional convolution, you use the convn function. convn takes as arguments a data array and a convolution kernel, both of which can be of any dimension, and returns an array whose dimension is the higher of the two input arrays’ dimensions. convn also takes a shape parameter argument that accepts the same values as in conv2 and filter2, and which has analogous effects in higher dimensions. One important application for the convn function is to filter image arrays that have multiple planes or frames. For example, suppose you have an array A containing five RGB images that you want to filter using a two-dimensional convolution kernel k. The image array is a four-dimensional array of size m-by-n-by-3-by-5. To filter this array with conv2, you would need to call the function 15 times, once for each combination of planes and frames, and assemble the results into a four-dimensional array. Using convn, you can filter the array in a single call: B = convn(A,k); For more information, see the reference entry for convn in Chapter 11. Using Predefined Filter Ty pes The function fspecial produces several kinds of predefined filters, in the form of computational molecules. After creating a filter with fspecial, you can www.MatlabSite.com 5-9 5 Linear Filtering and Filter Design apply it directly to your image data using filter2, or you can rotate it 180 degrees and use conv2 or convn. One simple filter fspecial can produce is an averaging filter. This type of filter computes the value of an output pixel by simply averaging the values of its neighboring pixels. The default size of the averaging filter fspecial creates is 3-by-3, but you can specify a different size. The value of each element is 1/length(h(:)). For example, a 5-by-5 averaging filter would be: 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 0.0400 Applying this filter to a pixel is equivalent to adding up the values of that pixel’s 5-by-5 neighborhood and dividing by 25. This has the effect of smoothing out local highlights and blurring edges in an image. This example illustrates applying a 5-by-5 averaging filter to an intensity image: I = imread('blood1.tif'); h = fspecial('average',5); I2 = uint8(round(filter2(h,I))); imshow(I) figure, imshow(I2) 5-10 www.MatlabSite.com Linear Filtering Note that the output from filter2 (and conv2 and convn) is always of class double. In the example above, the input image is of class uint8, so the output from filter2 consists of double-precision values in the range [0,255]. The call to the uint8 function converts the output to uint8; the data is not in the proper range for an image of class double. Another relatively simple filter fspecial can produce is a 3-by-3 Sobel filter, which is effective at detecting the horizontal edges of objects in an image: h = fspecial('sobel') h = 1 0 –1 2 0 –2 1 0 –1 Unlike an averaging filter, the Sobel filter produces values outside the range of the input data. If the input image is of class double, the output array may include values outside the range [0,1]; if the input image is of class uint8, the output array may include values outside the range [0,255]. To display the output as an image, you can use imshow and specify the data range, or you can use the mat2gray function to convert the values to the range [0,1]. Note that if the input image is of class uint8, you should not simply convert the output to uint8; if the output contains values outside the range [0,255], these values cannot be represent as 8-bit integers, so the result of the conversion may be unpredictable. You can, however, rescale the output and then convert it. For example: h = fspecial('sobel'); I2 = filter2(h,I); J = uint8(round(mat2gray(I2)*255)); The following example creates a Sobel filter and uses filter2 to apply the filter to the blood1 image. Notice that in the call to imshow, the intensity range is specified as an empty matrix ([]). This instructs imshow to display the minimum value in I2 as black, the maximum value as white, and values in www.MatlabSite.com 5-11 5 Linear Filtering and Filter Design between as intermediate shades of gray, thus enabling you to display the filter2 output without converting or rescaling it. I = imread('blood1.tif'); h = fspecial('sobel'); I2 = filter2(h,I); imshow(I2,[]) For a description of all the filter types fspecial provides, see the reference entry for fspecial in Chapter 11. 5-12 www.MatlabSite.com Filter Design Filter Design This section describes working in the frequency domain to design filters. Topics discussed include: • Finite Impulse Response (FIR) filters, the class of linear filter that the toolbox supports • The frequency transformation method, which transforms a one-dimensional FIR filter into a two-dimensional FIR filter • The frequency sampling method, which creates a filter based on a desired frequency response • The windowing method, which multiplies the ideal impulse response with a window function to generate the filter • Creating the desired frequency response matrix • Computing the frequency response of a filter This section assumes you are familiar with working in the frequency domain. This topic is discussed in many signal processing and image processing textbooks. N O TE Most of the design methods described in this section work by creating a two-dimensional filter from a one-dimensional filter or window created using functions from the Signal Processing Toolbox. Although this toolbox is not required, you may find it difficult to design filters in the Image Processing Toolbox if you do not have the Signal Processing Toolbox as well. FIR Filters The Image Processing Toolbox supports one class of linear filter, the two-dimensional Finite Impulse Response (FIR) filter. FIR filters have several characteristics that make them ideal for image processing in the MATLAB environment: • FIR filters are easy to represent as matrices of coefficients. • Two-dimensional FIR filters are natural extensions of one-dimensional FIR filters. www.MatlabSite.com 5-13 5 Linear Filtering and Filter Design • There are several well-known, reliable methods for FIR filter design. • FIR filters are easy to implement. • FIR filters can be designed to have linear phase, which helps prevent distortion. Another class of filter, the Infinite Impulse Response (IIR) filter, is not as suitable for image processing applications. It lacks the inherent stability and ease of design and implementation of the FIR filter. Therefore, this toolbox does not provide IIR filter support. Frequency Tra nsform a tion M ethod The frequency transformation method transforms a one-dimensional FIR filter into a two-dimensional FIR filter. The frequency transformation method preserves most of the characteristics of the one-dimensional filter, particularly the transition bandwidth and ripple characteristics. This method uses a transformation matrix, a set of elements that defines the frequency transformation. The toolbox function ftrans2 implements the frequency transformation method. This function’s default transformation matrix produces filters with nearly circular symmetry. By defining your own transformation matrix, you can obtain different symmetries. (See Jae S. Lim, Two-Dimensional Signal and Image Processing, 1990, for details.) The frequency transformation method generally produces very good results, as it is easier to design a one-dimensional filter with particular characteristics than a corresponding two-dimensional filter. For instance, the next example designs an optimal equiripple one-dimensional FIR filter and uses it to create a two-dimensional filter with similar characteristics. The shape of the 5-14 www.MatlabSite.com Filter Design one-dimensional frequency response is clearly evident in the two-dimensional response. b = remez(10,[0 0.4 0.6 1],[1 1 0 0]); h = ftrans2(b); [H,w] = freqz(b,1,64,'whole'); colormap(jet(64)) plot(w/pi–1,fftshift(abs(H))) figure, freqz2(h,[32 32]) Frequency Sa m pling M ethod The frequency sampling method creates a filter based on a desired frequency response. Given a matrix of points that defines the shape of the frequency response, this method creates a filter whose frequency response passes through those points. Frequency sampling places no constraints on the behavior of the frequency response between the given points; usually, the response ripples in these areas. The toolbox function fsamp2 implements frequency sampling design for two-dimensional FIR filters. fsamp2 returns a filter h with a frequency response that passes through the points in the input matrix Hd. The example below creates an 11-by-11 filter using fsamp2, and plots the frequency response of the resulting filter. (The freqz2 function in this example calculates the www.MatlabSite.com 5-15 5 Linear Filtering and Filter Design two-dimensional frequency response of a filter. For more information, see page 5-19.) Hd = zeros(11,11); Hd(4:8,4:8) = 1; [f1,f2] = freqspace(11,'meshgrid'); mesh(f1,f2,Hd), axis([–1 1 –1 1 0 1.2]), colormap(jet(64)) h = fsamp2(Hd); figure, freqz2(h,[32 32]), axis([–1 1 –1 1 0 1.2]) Notice the ripples in the actual frequency response, compared to the desired frequency response. These ripples are a fundamental problem with the frequency sampling design method. They occur wherever there are sharp transitions in the desired response. You can reduce the spatial extent of the ripples by using a larger filter. However, a larger filter does not reduce the height of the ripples, and requires more computation time for filtering. To achieve a smoother approximation to the desired frequency response, consider using the frequency transformation method or the windowing method. W indow ing M ethod The windowing method involves multiplying the ideal impulse response with a window function to generate a corresponding filter. Like the frequency sampling method, the windowing method produces a filter whose frequency response approximates a desired frequency response. The windowing method, however, tends to produce better results than the frequency sampling method. 5-16 www.MatlabSite.com Filter Design The toolbox provides two functions for window-based filter design, fwind1 and fwind2. fwind1 designs a two-dimensional filter by using a two-dimensional window that it creates from one or two one-dimensional windows that you specify. fwind2 designs a two-dimensional filter by using a specified two-dimensional window directly. fwind1 supports two different methods for making the two-dimensional windows it uses: • Transforming a single one-dimensional window to create a two-dimensional window that is nearly circularly symmetric, by using a process similar to rotation • Creating a rectangular, separable window from two one-dimensional windows, by computing their outer product The example below uses fwind1 to create an 11-by-11 filter from the desired frequency response Hd. Here, the hamming function from the Signal Processing Toolbox is used to create a one-dimensional window, which fwind1 then extends to a two-dimensional window. Hd = zeros(11,11); Hd(4:8,4:8) = 1; [f1,f2] = freqspace(11,'meshgrid'); mesh(f1,f2,Hd), axis([–1 1 –1 1 0 1.2]), colormap(jet(64)) h = fwind1(Hd,hamming(11)); figure, freqz2(h,[32 32]), axis([–1 1 –1 1 0 1.2]) www.MatlabSite.com 5-17 5 Linear Filtering and Filter Design Crea ting the Desired Frequency Response M a trix The filter design functions fsamp2, fwind1, and fwind2 all create filters based on a desired frequency response magnitude matrix. You can create an appropriate desired frequency response matrix using the freqspace function. freqspace returns correct, evenly spaced frequency values for any size response. If you create a desired frequency response matrix using frequency points other than those returned by freqspace, you may get unexpected results, such as nonlinear phase. For example, to create a circular ideal lowpass frequency response with cutoff at 0.5 use: [f1,f2] = freqspace(25,'meshgrid'); Hd = zeros(25,25); d = sqrt(f1.^2 + f2.^2) < 0.5; Hd(d) = 1; mesh(f1,f2,Hd) Note that for this frequency response, filters produced by fsamp2, fwind1, and fwind2 are real. This result is desirable for most image processing applications. To achieve this in general, the desired frequency response should be symmetric about the frequency origin (f1 = 0, f2 = 0). 5-18 www.MatlabSite.com Filter Design Com puting the Frequency Response of a Filter The freqz2 function computes the frequency response for a two-dimensional filter. With no output arguments, freqz2 creates a mesh plot of the frequency response. For example, consider this FIR filter: h =[0.1667 0.6667 0.1667 0.6667 –3.3333 0.6667 0.1667 0.6667 0.1667]; This command computes and displays the 64-by-64 point frequency response of h: freqz2(h) To obtain the frequency response matrix H and the frequency point vectors f1 and f2, use output arguments: [H,f1,f2] = freqz2(h); freqz2 normalizes the frequencies f1 and f2 so that the value 1.0 corresponds to half the sampling frequency, or π radians. For a simple m-by-n response, as shown above, freqz2 uses the two-dimensional fast Fourier transform function fft2. You can also specify vectors of arbitrary frequency points, but in this case freqz2 uses a slower algorithm. For more information about the fast Fourier transform and its application to linear filtering and filter design, see Chapter 6. www.MatlabSite.com 5-19 5 Linear Filtering and Filter Design 5-20 www.MatlabSite.com 6 Transforms Overview . . . . . . . . . . . . . . . . . . . . . 6-2 Fourier Transform . . . . . Definition . . . . . . . . . . The Discrete Fourier Transform Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-3 6-3 6-8 6-11 Discrete Cosine Transform . . . . . . . . . . . . . 6-15 The DCT Transform Matrix . . . . . . . . . . . . . . 6-16 The DCT and Image Compression . . . . . . . . . . . . 6-17 Radon Transform . . . . . . . . . . . . . . . . . 6-19 Using the Radon Transform to Detect Lines . . . . . . . . 6-22 The Inverse Radon Transform . . . . . . . . . . . . . 6-25 www.MatlabSite.com 6 Transforms O verview The usual mathematical representation of an image is a function of two spatial variables: f ( x, y ) . The value of the function at a particular location ( x, y ) represents the intensity of the image at that point. The term transform refers to an alternative mathematical representation of an image. For example, the Fourier transform is a representation of an image as a sum of complex exponentials of varying magnitudes, frequencies, and phases. This representation is useful in a broad range of applications, including (but not limited to) image analysis, restoration, and filtering. The discrete cosine transform (DCT) also represents an image as a sum of sinusoids of varying magnitudes and frequencies. The DCT is extremely useful for image compression; it is the basis of the widely used JPEG image compression algorithm. The Radon transform represents an image as a collection of projections along various directions. It is used in areas ranging from seismology to computer vision. This chapter defines each of these transforms, describes related toolbox functions, and shows examples of related image processing applications. 6-2 www.MatlabSite.com Fourier Transform Fourier Tra nsform The Fourier transform plays a critical role in a broad range of image processing applications, including enhancement, analysis, restoration, and compression. This section defines the Fourier transform and the related discrete Fourier transform, discusses the fast Fourier transform, and presents some sample applications. Definition If f ( m, n ) is a function of two discrete spatial variables m and n, then we define the two-dimensional Fourier transform of f ( m, n ) by the relationship: ∞ F ( ω 1, ω 2 ) = ∞ ∑ ∑ f ( m, n )e – jω 1 m e – jω 2 n m = –∞ n = –∞ The variables ω1 and ω2 are frequency variables; their units are radians per sample. F ( ω 1, ω 2 ) is often called the frequency-domain representation of f ( m, n ) . F ( ω 1, ω 2 ) is a complex-valued function that is periodic both in ω 1 and ω 2 , with period 2π . Because of the periodicity, usually only the range – π ≤ ω 1, ω 2 ≤ π is displayed. Note that F ( 0, 0 ) is the sum of all the values of f ( m, n ) . For this reason, F ( 0, 0 ) is often called the constant component or DC component of the Fourier transform. (DC stands for direct current; it is an electrical engineering term that refers to a constant-voltage power source, as opposed to a power source whose voltage varies sinusoidally.) The inverse two-dimensional Fourier transform is given by: 1 f ( m, n ) = -----2π π π ∫ω 1 = –π ∫ω 2 = –π F ( ω 1, ω 2 )e jω 1 m jω 2 n e dω 1 dω 2 Roughly speaking, this equation means that f ( m, n ) can be represented as a sum of an infinite number of complex exponentials (sinusoids) with different frequencies. The magnitude and phase of the contribution at the frequencies ( ω 1, ω 2 ) are given by F ( ω 1, ω 2 ) . www.MatlabSite.com 6-3 6 Transforms Example Consider a function f ( m, n ) that equals 1 within a rectangular region and 0 everywhere else. n f(m,n) m To simplify the diagram, f ( m, n ) is shown as a continuous function, even though the variables m and n are discrete. 6-4 www.MatlabSite.com Fourier Transform The magnitude of the Fourier transform, F ( ω 1, ω 2 ) , is shown here as a mesh plot: ω 2 (ve rti cal fre y) qu uenc enc y) freq ntal orizo ω 1 (h The peak at the center of the plot is F ( 0, 0 ) , which is the sum of all the values in f ( m, n ) . The plot also shows that F ( ω 1, ω 2 ) has more energy at high horizontal frequencies than at high vertical frequencies. This reflects the fact that horizontal cross sections of f ( m, n ) are narrow pulses, while vertical cross sections are broad pulses. Narrow pulses have more high-frequency content than broad pulses. www.MatlabSite.com 6-5 6 Transforms ω2 Another common way to visualize the Fourier transform is to display log F ( ω 1, ω 2 ) as an image, as in: ω1 Using the logarithm helps to bring out details of the Fourier transform in regions where F ( ω 1, ω 2 ) is very close to 0. 6-6 www.MatlabSite.com Fourier Transform Examples of the Fourier transform for other simple shapes are shown below. www.MatlabSite.com 6-7 6 Transforms The Discrete Fourier Tra nsform Working with the Fourier transform on a computer usually involves a form of the transform known as the discrete Fourier transform (DFT). There are two principal reasons for using this form: • The input and output of the DFT are both discrete, which makes it convenient for computer manipulations. • There is a fast algorithm for computing the DFT known as the fast Fourier transform (FFT). The DFT is usually defined for a discrete function f ( m, n ) that is nonzero only over the finite region 0 ≤ m ≤ M – 1 and 0 ≤ n ≤ N – 1 . The two-dimensional M-by-N DFT and inverse M-by-N DFT relationships are given by: M–1 N–1 F ( p, q ) = ∑ ∑ f ( m, n )e – j ( 2π ⁄ M ) pm – j ( 2π ⁄ N )qn e m=0 n=0 1 f ( m, n ) = ----------MN M–1 N–1 ∑ ∑ p=0 F ( p, q )e j ( 2π ⁄ M ) pm j ( 2π ⁄ N )qn q=0 e p = 0, 1, …, M – 1 q = 0, 1, … , N – 1 m = 0, 1, … , M – 1 n = 0, 1, …, N – 1 The values F ( p, q ) are called the DFT coefficients of f ( m, n ) . In particular, the value F ( 0, 0 ) is called the DC coefficient. (Note that matrix indices in MATLAB always start at 1 rather than 0; therefore, the matrix elements f(1,1) and F(1,1) correspond to the mathematical quantities f ( 0, 0 ) and F ( 0, 0 ) , respectively.) The MATLAB functions fft, fft2, and fftn implement the fast Fourier transform algorithm for computing the one-dimensional DFT, two-dimensional DFT, and N-dimensional DFT, respectively. The functions ifft, ifft2, and ifftn compute the inverse DFT. Relationship to the Fourier Transform The DFT coefficients F ( p, q ) are samples of the Fourier transform F ( ω 1, ω 2 ) : 6-8 www.MatlabSite.com Fourier Transform F ( p, q ) = F ( ω 1, ω 2 ) ω 1 = 2πp ⁄ M ω 2 = 2πq ⁄ N p = 0, 1, …, M – 1 q = 0, 1, …, N – 1 Example Let’s construct a matrix f that is similar to the function f(m,n) in the example on page 6-4. f = zeros(30,30); f(5:24,13:17) = 1; imshow(f,'notruesize') www.MatlabSite.com 6-9 6 Transforms Compute and visualize the 30-by-30 DFT of f with these commands: F = fft2(f); F2 = log(abs(F)); imshow(F2,[–1 5],'notruesize'); colormap(jet); colorbar This plot differs from the Fourier transform displayed on page 6-6. First, the resolution is much lower. Second, the DC coefficient is displayed in the upper-left corner instead of the traditional location in the center. The resolution can be increased by zero-padding f when computing its DFT. The zero-padding and DFT computation can be performed in a single step with this command: F = fft2(f,256,256); 6-10 www.MatlabSite.com Fourier Transform This command zero-pads f to be 256-by-256 before computing the DFT. The result is that the DCT coefficients F sample the Fourier transform much more finely: imshow(log(abs(F)),[–1 5]); colormap(jet); colorbar The DC coefficient, however, is still displayed in the upper-left corner rather than the center. You can fix this problem by using the function fftshift, which swaps the quadrants of F so that the DC coefficient is in the center. F = fft2(f,256,256); F2 = fftshift(F); imshow(log(abs(F2)),[–1 5]); colormap(jet); colorbar The resulting plot is identical to the one on page 6-6. Applica tions This section presents a few of the many image processing-related applications of the Fourier transform. Frequency Response of Linear Filters The Fourier transform of the impulse response of a linear filter gives the frequency response of the filter. The function freqz2 computes and displays a filter’s frequency response. The frequency response of the Gaussian www.MatlabSite.com 6-11 6 Transforms convolution kernel shows that this filter passes low frequencies and attenuates high frequencies. h = fspecial('gaussian'); freqz2(h) See Chapter 5 for more information about linear filtering, filter design, and frequency responses. Fast Convolution A key property of the Fourier transform is that the multiplication of two Fourier transforms corresponds to the convolution of the associated spatial functions. This property, together with the fast Fourier transform, forms the basis for a fast convolution algorithm. Suppose that A is an M-by-N matrix and B is a P-by-Q matrix. The convolution of A and B can be computed using the following steps: 1 Zero-pad A and B so that they are at least (M+P–1)-by-(N+Q–1). (Often A and B are zero-padded to a size that is a power of 2 because fft2 is fastest for these sizes.) 2 Compute the two-dimensional DFT of A and B using fft2. 6-12 www.MatlabSite.com Fourier Transform 3 Multiply the two DFTs together. 4 Using ifft2, compute the inverse two-dimensional DFT of the result from step 3. For example: A = magic(3); B = ones(3); A(8,8) = 0; B(8,8) = 0; C = ifft2(fft2(A).*fft2(B)); C = C(1:5,1:5); C = real(C) % remove imaginary % zero–pad A to be 8–by–8 % zero–pad B to be 8–by–8 % extract the nonzero portion part caused by roundoff error C = 8.0000 11.0000 15.0000 7.0000 4.0000 9.0000 17.0000 30.0000 21.0000 13.0000 15.0000 30.0000 45.0000 30.0000 15.0000 7.0000 19.0000 30.0000 23.0000 11.0000 6.0000 13.0000 15.0000 9.0000 2.0000 The FFT-based convolution method is most often used for large inputs. For small inputs it is generally faster to use filter2 or conv2. Locating Image Features The Fourier transform can also be used to perform correlation, which is closely related to convolution. Correlation can be used to locate features within an image; in this context correlation is often called template matching. For instance, suppose you want to locate occurrences of the letter “a” in the image www.MatlabSite.com 6-13 6 Transforms text.tif. The figure below shows the image and the template you want to match: The correlation of the image of the letter “a” with the larger image can be computed by first rotating the image of “a” by 180o and then using the FFT-based convolution technique described above. The left image below is the result of the correlation; bright peaks correspond to occurrences of the letter. The locations of these peaks are indicated by the white spots in the thresholded correlation image shown on the right. 6-14 www.MatlabSite.com Discrete C osine Transform Discrete Cosine Tra nsform The dct2 function in the Image Processing Toolbox computes the two-dimensional discrete cosine transform (DCT) of an image. The DCT has the property that, for a typical image, most of the visually significant information about the image is concentrated in just a few coefficients of the DCT. For this reason, the DCT is often used in image compression applications. For example, the DCT is at the heart of the the international standard lossy image compression algorithm known as JPEG. (The name comes from the working group that developed the standard: the Joint Photographic Experts Group.) The two-dimensional DCT of an M-by-N matrix A is defined as follows: M–1 N–1 B pq = α p α q ∑ ∑ m=0 n=0 π ( 2m + 1 ) p π ( 2n + 1 )q A mn cos ------------------------------- cos ----------------------------- , 2M 2N  1 ⁄ M, p = 0 αp =   2 ⁄ M, 1 ≤ p ≤ M – 1 0≤ p≤M–1 0≤q≤N–1 1 ⁄ N , q = 0 αq =   2 ⁄ N, 1 ≤ q ≤ N – 1 The values B pq are called the DCT coefficients of A. (Note that matrix indices in MATLAB always start at 1 rather than 0; therefore, the MATLAB matrix elements A(1,1) and B(1,1) correspond to the mathematical quantities A 00 and B 00 , respectively.) The DCT is an invertible transform, and its inverse is given by: M–1 N–1 A mn = ∑ ∑ p=0 q=0 π ( 2m + 1 ) p π ( 2n + 1 )q α p α q B pq cos ------------------------------- cos ----------------------------- , 2M 2N  1 ⁄ M, p = 0 αp =   2 ⁄ M, 1 ≤ p ≤ M – 1 0≤m≤M–1 0≤n≤N–1 1 ⁄ N , q = 0 αq =   2 ⁄ N, 1 ≤ q ≤ N – 1 The inverse DCT equation can be interpreted as meaning that any M-by-N matrix A can be written as a sum of MN functions of the form: www.MatlabSite.com 6-15 6 Transforms π ( 2m + 1 ) p π ( 2n + 1 )q α p α q cos ------------------------------- cos ----------------------------- , 2M 2N 0≤ p≤M–1 0≤q≤N–1 These functions are called the basis functions of the DCT. The DCT coefficients B pq , then, can be regarded as the weights applied to each basis function. For 8-by-8 matrices, the 64 basis functions are illustrated by this image: Horizontal frequencies increase from left to right, and vertical frequencies increase from top to bottom. The constant-valued basis function at the upper left is often called the DC basis function, and the corresponding DCT coefficient B00 is often called the DC coefficient. The DCT Tra nsform M a trix The Image Processing Toolbox offers two different ways to compute the DCT. The first method is to use the function dct2. dct2 uses an FFT-based algorithm for speedy computation with large inputs. For small square inputs, such as 8-by-8 or 16-by-16, it may be more efficient to use the DCT transform matrix, which is returned by the function dctmtx. The M-by-M transform matrix T is given by: 6-16 www.MatlabSite.com Discrete C osine Transform T pq  1  --------- =  M  2 π ( 2q + 1 ) p  ------ cos ---------------------------2M M  p = 0, 0 ≤ q ≤ M – 1 1 ≤ p ≤ M – 1, 0 ≤ q ≤ M – 1 For an M-by-M matrix A, T*A is an M-by-M matrix whose columns contain the one-dimensional DCT of the columns of A. The two-dimensional DCT of A can be computed as B=T*A*T'. Since T is a real orthonormal matrix, its inverse is the same as its transpose. Therefore, the inverse two-dimensional DCT of B is given by T'*B*T. The DCT a nd Im a ge Com pression In the JPEG image compression algorithm, the input image is divided into 8-by-8 or 16-by-16 blocks, and the two-dimensional DCT is computed for each block. The DCT coefficients are then quantized, coded, and transmitted. The JPEG receiver (or JPEG file reader) decodes the quantized DCT coefficients, computes the inverse two-dimensional DCT of each block, and then puts the blocks back together into a single image. For typical images, many of the DCT coefficients have values close to zero; these coefficients can be discarded without seriously affecting the quality of the reconstructed image. The example code below computes the two-dimensional DCT of 8-by-8 blocks in the input image; discards (sets to zero) all but 10 of the 64 DCT coefficients in www.MatlabSite.com 6-17 6 Transforms each block; and then reconstructs the image using the two-dimensional inverse DCT of each block. The transform matrix computation method is used. I = imread('cameraman.tif'); I = double(I)/255; T = dctmtx(8); B = blkproc(I,[8 8],'P1*x*P2',T,T'); mask = [1 1 1 1 0 0 0 0 1 1 1 0 0 0 0 0 1 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]; B2 = blkproc(B,[8 8],'P1.*x',mask); I2 = blkproc(B2,[8 8],'P1*x*P2',T',T); imshow(I), figure, imshow(I2) Although there is some loss of quality in the reconstructed image, it is clearly recognizable, even though almost 85% of the DCT coefficients were discarded. To experiment with discarding more or fewer coefficients, and to apply this technique to other images, try running the demo function dctdemo. 6-18 www.MatlabSite.com Radon Transform Ra don Tra nsform The radon function in the Image Processing Toolbox computes projections of an image matrix along specified directions. A projection of a two-dimensional function f(x,y) is a line integral in a certain direction. For example, the line integral of f(x,y) in the vertical direction is the projection of f(x,y) onto the x-axis; the line integral in the horizontal direction is the projection of f(x,y) onto the y-axis. This figure shows horizontal and vertical projections for a simple two-dimensional function. f(x,y) x Projection onto the y-axis y Projection onto the x-axis Projections can be computed along any angle θ. In general, the Radon transform of f(x,y) is the line integral of f parallel to the y′ axis: R θ ( x′ ) = ∞ ∫–∞ f (x′ cos θ – y′ sin θ, x′ sin θ + y′ cos θ) d y′ where x′ = cos θ sin θ x y′ – sin θ cos θ y www.MatlabSite.com 6-19 6 Transforms This figure illustrates the geometry of the Radon transform. y′ y x′ f(x,y) θ x x′ Rθ(x′) This command computes the Radon transform of I for the angles specified in the vector theta: [R,xp] = radon(I,theta); The columns of R contain the Radon transform for each angle in theta. xp contains the corresponding coordinates along the x′-axis. The “center pixel” of I is defined to be floor((size(I)+1)/2); this is the pixel on the x′-axis corresponding to x′ = 0 . 6-20 www.MatlabSite.com Radon Transform The commands below compute and plot the Radon transform at 0˚ and 45˚ of an image containing a single square object. I = zeros(100,100); I(25:75, 25:75) = 1; imshow(I) [R,xp] = radon(I,[0 45]); figure; plot(xp,R(:,1)); title('R_{0^o} (x\prime)') figure; plot(xp,R(:,2)); title('R_{45^o} (x\prime)') R 0° ( x′ ) www.MatlabSite.com R 45° ( x′ ) 6-21 6 Transforms The Radon transform for a large number of angles is often displayed as an image. In this example, the Radon transform for the square image is computed at angles from 0˚ to 180˚, in 1˚ increments: theta = 0:180; [R,xp] = radon(I,theta); imagesc(theta,xp,R); xlabel('\theta (degrees)'); ylabel('x'''); colormap(hot); colorbar R θ ( x′ ) Using the Ra don Tra nsform to Detect Lines The Radon transform is closely related to a common computer vision operation known as the Hough transform. You can use the radon function to implement a form of the Hough transform used to detect straight lines. The steps are: 6-22 www.MatlabSite.com Radon Transform 1 Compute a binary edge image using the edge function. I = imread('ic.tif'); BW = edge(I); imshow(I) figure, imshow(BW) 2 Compute the Radon transform of the edge image. www.MatlabSite.com 6-23 6 Transforms theta = 0:179; [R,xp] = radon(BW,theta); imagesc(theta, xp, R); colormap(hot); xlabel('\theta (degrees)'); ylabel('x'''); colorbar R θ ( x′ ) 3 Find the locations of strong peaks in the Radon transform matrix. The locations of these peaks correspond to the location of straight lines in the original image. In this example, the strongest peak in R corresponds to θ = 94° and x′ = – 101 . The line perpendicular to that angle and located at x′ = – 101 is 6-24 www.MatlabSite.com Radon Transform shown below, superimposed in red on the original image. The Radon transform geometry is shown in black. x′ θ = 94˚ x x′ = –101 Notice that the other strong lines parallel to the red line also appear as peaks at θ = 94° in the transform. Also, the lines perpendicular to this line appear as peaks at θ = 4° . The Inverse Ra don Tra nsform The iradon function in the toolbox uses the inverse Radon transform to reconstruct images from projection data, such as the output from the radon function. The inverse Radon transform is commonly used in tomography applications. For more information about iradon, see the reference entry for this function in Chapter 11. Also, see the entry for the phantom function, which you can use to generate images for testing the Radon and inverse Radon transforms. www.MatlabSite.com 6-25 6 Transforms 6-26 www.MatlabSite.com 7 Analyzing and Enhancing Images Overview . . . . . . . . . . . . . . . . . . . . . 7-2 Pixel Values and Statistics Pixel Selection . . . . . . . Intensity Profile . . . . . . Image Contours . . . . . . Image Histogram . . . . . Summary Statistics . . . . . Feature Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-3 7-3 7-4 7-7 7-8 7-9 7-9 Image Analysis . . . . . . . . . . . . . . . . . . . 7-10 Edge Detection . . . . . . . . . . . . . . . . . . . 7-10 Quadtree Decomposition . . . . . . . . . . . . . . . . 7-11 Image Enhancement . . . . . . . . . . . . . . . . 7-14 Intensity Adjustment . . . . . . . . . . . . . . . . . 7-14 Noise Removal . . . . . . . . . . . . . . . . . . . . 7-20 www.MatlabSite.com 7 Analyzing and Enhancing Images O verview The Image Processing Toolbox supports a range of standard image processing operations for analyzing and enhancing images. Its functions simplify several categories of tasks, including: • Obtaining pixel values and statistics, which are numerical summaries of data in an image • Analyzing images to extract information about their essential structure • Enhancing images to make certain features easier to see or to reduce noise This section describes specific operations within each category, and shows how to implement each kind of operation using toolbox functions. 7-2 www.MatlabSite.com Pixel Values and Statistics Pix el Va lues a nd Sta tistics The Image Processing Toolbox provides several functions that return information about the data values that make up an image. These functions return information about image data in various forms, including: • The data values for selected pixels (impixel, pixval) • The data values along a path in an image (improfile) • A contour plot of the image data (imcontour) • A histogram of the image data (imhist) • Summary statistics for the image data (mean2, std2, corr2) • Feature measurements for image regions (imfeature) Pix el Selection The impixel function returns the data values for a selected pixel or set of pixels. You can supply the coordinates of the pixels as input arguments, or you can select pixels using a mouse. If you call impixel with no input arguments, the cursor changes to a cross hair when it is over the image. You can then click on the pixels of interest; impixel displays a small star over each pixel you select. When you are done selecting pixels, press Return. impixel returns the color values for the selected pixels, and removes the stars. impixel works with indexed, intensity, and RGB images. impixel always returns pixel values as RGB triplets, regardless of the image type: • For an RGB image, impixel returns the actual data for the pixel. The values are either uint8 integers or double floating-point numbers, depending on the class of the image array. • For an indexed image, impixel returns the RGB triplet stored in the row of the colormap that the pixel value points to. The values are double floating-point numbers. • For an intensity image, impixel returns the intensity value as an RGB triplet, where R=G=B. The values are either uint8 integers or double floating-point numbers, depending on the class of the image array. www.MatlabSite.com 7-3 7 Analyzing and Enhancing Images In this example, you call impixel and click on three points in the displayed image, and then press Return. imshow canoe.tif impixel * * * ans = 0.1294 0.5176 0.7765 0.1294 0 0.6118 0.1294 0 0.4196 Notice that the second pixel, which is part of the canoe, is pure red; its green and blue values are both 0. The toolbox also provides the pixval function, which interactively displays the data values for pixels as you move the cursor over the image. pixval can also display the Euclidean distance between two pixels. For more information about pixval, see the reference entry for this function in Chapter 11. Intensity Profile The improfile function calculates and plots the intensity values along a line segment or a multiline path in an image. You can supply the coordinates of the line segments as input arguments, or you can define the desired path using a mouse. In either case, improfile uses interpolation to determine the values of equally spaced points along the path. (By default, improfile uses nearest neighbor interpolation, but you can specify a different method. See Chapter 3 7-4 www.MatlabSite.com Pixel Values and Statistics for a discussion of interpolation.) improfile works best with intensity and RGB images. (For indexed images, it displays the index values from the image matrix, not the intensity values from the colormap.) For a single line segment, improfile plots the intensity values in a two-dimensional view. For a multiline path, improfile plots the intensity values in a three-dimensional view. If you call improfile with no arguments, the cursor changes to a cross hair when it is over the image. You can then specify line segments by clicking on the endpoints; improfile draws a line between each two consecutive points you select. When you finish specifying the path, press Return. improfile displays the plot in a new figure. In this example, you call improfile and specify a single line with the mouse. The line is shown in red, and is drawn from left to right. imshow debye1.tif improfile www.MatlabSite.com 7-5 7 Analyzing and Enhancing Images improfile displays a plot of the data along the line: Notice the peaks and valleys and how they correspond to the light and dark bands in the image. The example below shows how improfile works with an RGB image. The red line is drawn from top to bottom. imshow flowers.tif improfile 7-6 www.MatlabSite.com Pixel Values and Statistics improfile displays a plot with separate lines for the red, green, and blue intensities: Notice how the lines correspond to the colors in the image. For example, the central region of the plot shows high intensities of green and red, while the blue intensity is 0. These are the values for the yellow flower. Im a ge Contours You can use the toolbox function imcontour to display a contour plot of the data in an intensity image. This function is similar to the contour function in MATLAB, but it automatically sets up the axes so their orientation and aspect ratio match the image. www.MatlabSite.com 7-7 7 Analyzing and Enhancing Images This example displays a grayscale image of grains of rice and a contour plot of the image data. I = imread('rice.tif'); imshow(I) figure, imcontour(I) You can use the clabel function to label the levels of the contours. See the description of clabel in the online MATLAB Function Reference for details. Im a ge H istogra m An image histogram is a chart that shows the distribution of intensities in an indexed or intensity image. The image histogram function imhist creates this plot by making n equally spaced bins, each representing a range of data values. It then calculates the number of pixels within each range. For example, the 7-8 www.MatlabSite.com Pixel Values and Statistics commands below display an image of grains of rice, and a histogram based on 64 bins. I = imread('rice.tif'); imshow(I) figure, imhist(I,64) The histogram shows a peak at around 100, due to the dark background in the image. For information about how to modify an image by changing the distribution of its histogram, see “Intensity Adjustment” on page 7-14. Sum m a r y Sta tistics You can compute standard image statistics using the mean2, std2, and corr2 functions. mean2 and std2 compute the mean and standard deviation of the elements of a matrix. corr2 computes the correlation coefficient between two matrices of the same size. These functions are two-dimensional versions of the mean, std, and corrcoef functions described in the online MATLAB Function Reference. Fea ture M ea surem ent You can use the imfeature function to compute feature measurements for image regions. For example, imfeature can measure such features as the area, center of mass, and bounding box for a region you specify. See the reference entry for imfeature in Chapter 11 for more information. www.MatlabSite.com 7-9 7 Analyzing and Enhancing Images Ima ge Ana lysis Image analysis techniques return information about the structure of an image. This section describes toolbox functions that you can use for these image analysis techniques: • Edge detection • Quadtree decomposition The functions described in this section work only with grayscale intensity images. Edge Detection You can use the edge function to detect edges, which are those places in an image that correspond to object boundaries. To find edges, this function looks for places in the image where the intensity changes rapidly, using one of these two criteria: • Places where the first derivative of the intensity is larger in magnitude than some threshold • Places where the second derivative of the intensity has a zero crossing edge provides a number of derivative estimators, each of which implements one of the definitions above. For some of these estimators, you can specify whether the operation should be sensitive to horizontal or vertical edges, or both. edge returns a binary image containing 1’s where edges are found and 0’s elsewhere. 7-10 www.MatlabSite.com Image Analysis The example below uses the Sobel method to detect the edges in an image of blood cells. By default, the operation is sensitive to both horizontal and vertical edges. I = imread('blood1.tif'); BW = edge(I,'sobel'); imshow(I) figure, imshow(BW) For an interactive demonstration of edge detection, try running edgedemo. Q ua dtree Decom position Quadtree decomposition is an analysis technique that involves subdividing an image into blocks that are more homogeneous than the image itself. This technique reveals information about the structure of the image. It is also useful as the first step in adaptive compression algorithms. You can perform quadtree decomposition using the qtdecomp function. This function works by dividing a square image into four equal-sized square blocks, and then testing each block to see if it meets some criterion of homogeneity (e.g., if all of the pixels in the block are within a specific dynamic range). If a block meets the criterion, it is not divided any further. If it does not meet the criterion, it is subdivided again into four blocks, and the test criterion is applied to those blocks. This process is repeated iteratively until each block meets the criterion. The result may have blocks of several different sizes. For example, suppose you want to perform quadtree decomposition on a 128-by-128 grayscale image. The first step is to divide the image into four www.MatlabSite.com 7-11 7 Analyzing and Enhancing Images 64-by-64 blocks. You then apply the test criterion to each block; for example, the criterion might be: max(block(:)) – min(block(:)) <= 0.2 If one of the blocks meets this criterion, it is not divided any further; it is 64-by-64 in the final decomposition. If a block does not meet the criterion, it is then divided into four 32-by-32 blocks, and the test is then applied to each of these blocks. The blocks that fail to meet the criterion are then divided into four 16-by-16 blocks, and so on, until all blocks “pass.” Some of the blocks may be as small as 1-by-1, unless you specify otherwise. The call to qtdecomp for this example would be: S = qtdecomp(I,0.2) S is returned as a sparse matrix whose nonzero elements represent the upper-left corners of the blocks; the value of each nonzero element indicates the block size. S is the same size as I. Note that the threshold value is specified as a value between 0 and 1, even if I is of class uint8. If I is uint8, the threshold value you supply is multiplied by 255 to determine the actual threshold to use. The example below shows an image and a representation of its quadtree decomposition. Each black square represents a homogeneous block, and the white lines represent the boundaries between blocks. Notice how the blocks are smaller in areas corresponding to large changes in intensity in the image. You can also supply qtdecomp with a function (rather than a threshold value) for deciding whether to split blocks; for example, you might base the decision 7-12 www.MatlabSite.com Image Analysis on the variance of the block. See the reference entry for qtdecomp in Chapter 11 for more information. For an interactive demonstration of quadtree decomposition, try running qtdemo. www.MatlabSite.com 7-13 7 Analyzing and Enhancing Images Ima ge Enha ncement Image enhancement techniques are used to improve an image, where “improve” is sometimes defined objectively (e.g., increase the signal-to-noise ratio), and sometimes subjectively (e.g., make certain features easier to see by modifying the colors or intensities). This section discusses these image enhancement techniques: • Intensity adjustment • Noise removal The functions described in this section apply primarily to grayscale intensity images. However, some of these functions can be applied to color images as well. For information about how these functions work with color images, see the reference entries for the individual functions in Chapter 11. Intensity Adjustm ent Intensity adjustment is a technique for mapping an image’s intensity values to a new range. For example, look at the rice image and its histogram on page 7-9. Notice that the image has rather low contrast, and that the histogram indicates no values below 40 or above 255. If you remap the data values to fill the entire intensity range [0,255], you can increase the contrast of the image. You can do this kind of adjustment with the imadjust function. For example, this code performs the adjustment described above: I = imread('rice.tif'); J = imadjust(I,[0.15 0.9],[0 1]); Notice that the intensities are specified as values between 0 and 1, even if I is of class uint8. If I is uint8, the values you supply are multiplied by 255 to determine the actual values to use. 7-14 www.MatlabSite.com Image Enhancement Now display the adjusted image and its histogram: imshow(J) figure, imhist(J,64) Notice the increased contrast in the image, and that the histogram now fills the entire range. Similarly, you can decrease the contrast of an image by narrowing the range of the data, as in this call: J = imadjust(I,[0 1],[0.3 0.8]); The general syntax is: J = imadjust(I,[low high],[bottom top]) where low and high are the intensities in the input image, which are mapped to bottom and top in the output image. In addition to increasing or decreasing contrast, you can perform a wide variety of other image enhancements with imadjust. In the example below, the man’s coat is too dark to reveal any detail. The call to imadjust maps the range [0,51] in the original image to [128,255] in the output image. This brightens the www.MatlabSite.com 7-15 7 Analyzing and Enhancing Images image considerably, and also widens the dynamic range of the dark portions of the original image, making it much easier to see the details in the coat. I = imread('cameraman.tif'); J = imadjust(I,[0 0.2],[0.5 1]); imshow(I) figure, imshow(J) Notice that this operation results in much of the image being washed out. This is because all values above 51 in the original image get mapped to 255 in the adjusted image. Gamma Correction imadjust maps low to bottom, and high to top. By default, the values between low and high are mapped linearly to values between bottom and top. For example, the value halfway between low and high corresponds to the value halfway between bottom and top. imadjust can accept an additional argument which specifies the gamma correction factor. Depending on the value of gamma, the mapping between values in the input and output images may be nonlinear. For example, the value halfway between low and high may map to a value either greater than or less than the value halfway between bottom and top. Gamma can be any value between 0 and infinity. If gamma is 1 (the default), the mapping is linear. If gamma is less than 1, the mapping is weighted toward higher (brighter) output values. If gamma is greater than than 1, the mapping is weighted toward lower (darker) output values. 7-16 www.MatlabSite.com Image Enhancement The figure below illustrates this relationship. The three transformation curves show how values are mapped when gamma is less than, equal to, and greater than 1. (In each graph, the x-axis represents the intensity values in the input image, and the y-axis represents the intensity values in the output image.) γ<1 γ=1 γ>1 top top top bottom bottom bottom low high low high low high The example below illustrates gamma correction. Notice that in the call to imadjust, the data ranges of the input and output images are specified as empty matrices. When you specify an empty matrix, imadjust uses the default range of [0,1]. In the example, both ranges are left empty; this means that gamma correction is applied without any other adjustment of the data. [X,map] = imread('forest.tif') I = ind2gray(X,map); J = imadjust(I,[],[],0.5); imshow(I) figure, imshow(J) www.MatlabSite.com 7-17 7 Analyzing and Enhancing Images Histogram Equalization The process of adjusting intensity values can be done automatically by the histeq function. histeq performs histogram equalization, which involves transforming the intensity values so that the histogram of the output image approximately matches a specified histogram. (By default, histeq tries to match a flat histogram with 64 bins, but you can specify a different histogram instead; see the reference entry for histeq in Chapter 11.) This example illustrates using histeq to adjust an intensity image. The original image has low contrast, with most values in the middle of the intensity range. histeq produces an output image having values evenly distributed throughout the range. I = imread('pout.tif'); J = histeq(I); imshow(I) figure, imshow(J) 7-18 www.MatlabSite.com Image Enhancement The example below shows the histograms for the two images: imhist(I) figure, imhist(J) histeq can return an additional 1-by-256 vector that shows, for each possible input value, the resulting output value. (The values in this vector are in the www.MatlabSite.com 7-19 7 Analyzing and Enhancing Images range [0,1], even if the image is of class uint8.) You can plot this data to get the tranformation curve. For example: I = imread('pout.tif'); [J,T] = histeq(I); plot((0:255)/255,T); Notice how this curve reflects the histograms in the previous figure, with the input values being mostly between 0.3 and 0.6, while the output values are distributed evenly between 0 and 1. For an interactive demonstration of intensity adjustment, try running imadjdemo. N oise Rem ova l Digital images are prone to a variety of types of noise. There are several ways that noise can be introduced into an image, depending on how the image is created. For example: • If the image is scanned from a photograph made on film, the film grain is a source of noise. Noise can also be the result of damage to the film, or be introduced by the scanner itself. • If the image is acquired directly in a digital format, the mechanism for gathering the data (such as a CCD detector) can introduce noise. • Electronic transmission of image data can introduce noise. 7-20 www.MatlabSite.com Image Enhancement The toolbox provides a number of different ways to remove or reduce noise in an image. Different methods are better for different kinds of noise. The methods available include: • Linear filtering • Median filtering • Adaptive filtering Also, in order to simulate the effects of some of the problems listed above, the toolbox provides the imnoise function, which you can use to add various types of noise to an image. The examples in this section use this function. Linear Filtering You can use linear filtering to remove certain types of noise. Certain filters, such as averaging or Gaussian filters, are appropriate for this purpose. For example, an averaging filter is useful for removing grain noise from a photograph. Because each pixel gets set to the average of the pixels in its neighborhood, local variations caused by grain are reduced. For more information about linear filtering, see Chapter 5. M edian Filtering Median filtering is similar to using an averaging filter, in that each output pixel is set to an “average” of the pixel values in the neighborhood of the corresponding input pixel. However, with median filtering, the value of an output pixel is determined by the median of the neighborhood pixels, rather than the mean. The median is much less sensitive than the mean to extreme values (called outliers). Median filtering is therefore better able to remove these outliers without reducing the sharpness of the image. The medfilt2 function implements median filtering. The example below compares using an averaging filter and medfilt2 to remove salt and pepper noise. This type of noise consists of random pixels being set to black or white (the extremes of the data range). In both cases the size of the neighborhood used for filtering is 3-by-3. www.MatlabSite.com 7-21 7 Analyzing and Enhancing Images First, read in the image and add noise to it. I = imread('eight.tif'); J = imnoise(I,'salt & pepper',0.02); imshow(I) figure, imshow(J) Now filter the noisy image and display the results. Notice that medfilt2 does a better job of removing noise, with less blurring of edges. K = filter2(fspecial('average',3),J)/255; L = medfilt2(J,[3 3]); imshow(K) figure, imshow(L) Median filtering is a specific case of order-statistic filtering. For information about order-statistic filtering, see the reference entry for the ordfilt2 function in Chapter 11. 7-22 www.MatlabSite.com Image Enhancement Adaptive Filtering The wiener2 function applies a Wiener filter (a type of linear filter) to an image adaptively, tailoring itself to the local image variance. Where the variance is large, wiener2 performs little smoothing. Where the variance is small, wiener2 performs more smoothing. This approach often produces better results than linear filtering. The adaptive filter is more selective than a comparable linear filter, preserving edges and other high frequency parts of an image. In addition, there are no design tasks; the wiener2 function handles all preliminary computations, and implements the filter for an input image. wiener2, however, does require more computation time than linear filtering. wiener2 work best when the noise is constant-power (“white”) additive noise, such as Gaussian noise. This example applies wiener2 to an image of Saturn that has had Gaussian noise added: I = imread('saturn.tif'); J = imnoise(I,'gaussian',0,0.005); K = wiener2(J,[5 5]); imshow(J) figure, imshow(K) For an interactive demonstration of filtering to remove noise, try running nrfiltdemo. www.MatlabSite.com 7-23 7 Analyzing and Enhancing Images 7-24 www.MatlabSite.com 8 Binary Image Operations Overview . . . . . . Neighborhoods . . . . . Padding of Borders . . . Displaying Binary Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-2 8-2 8-2 8-3 Morphological Operations . . . . . . . . . . . . . . 8-4 Dilation and Erosion . . . . . . . . . . . . . . . . . 8-4 Related Operations . . . . . . . . . . . . . . . . . . 8-7 Object-Based Operations . . . 4- and 8-Connected Neighborhoods Perimeter Determination . . . . Flood Fill . . . . . . . . . . . Connected-Components Labeling . Object Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-10 8-10 8-12 8-13 8-15 8-16 Feature Measurement . . . . . . . . . . . . . . . 8-18 Image Area . . . . . . . . . . . . . . . . . . . . . 8-18 Euler Number . . . . . . . . . . . . . . . . . . . . 8-19 Lookup-Table Operations . . . . . . . . . . . . . . 8-20 www.MatlabSite.com 8 Binary Image O perations O verview A binary image is an image in which each pixel assumes one of only two discrete values. Essentially, these two values correspond to on and off. Looking at an image in this way makes it easier to distinguish structural features. For example, in a binary image, it is easy to distinguish objects from the background. In the Image Processing Toolbox, a binary image is stored as a two-dimensional matrix of 0’s (which represent off pixels) and 1’s (which represent on pixels). The on pixels are the foreground of the image, and the off pixels are the background. Binary image operations return information about the form or structure of binary images only. To perform these operations on another type of image, you must first convert it to binary (using, for example, the im2bw function). N eighborhoods Most binary image algorithms work with groups of pixels called neighborhoods. A pixel’s neighborhood is some set of pixels, defined by their locations relative to that pixel. The neighborhood can include or omit the pixel itself, and the pixels included in the neighborhood are not necessarily adjacent to the pixel of interest. Different types of neighborhoods are used for different binary operations. Pa dding of Borders If a pixel is near the border of an image, some of the pixels in its neighborhood may be missing. For example, if the neighborhood is defined to include the pixel directly above the pixel of interest, then a pixel in the top row of an image will be missing this neighbor. In order to determine how to process these pixels, the binary image functions pad the borders of the image, usually with 0’s. In other words, these functions process the border pixels by assuming that the image is surrounded by additional rows and columns of 0’s. These rows and columns do not become part of the output image and are used only as parts of the neighborhoods of the actual pixels in the image. However, the padding can in some cases produce border effects, in which the regions near the borders of the output image do not appear to be homogeneous with the rest of the image. 8-2 www.MatlabSite.com O verview Displa y ing Bina r y Im a ges When you display a binary image with imshow, by default the foreground (i.e., the on pixels) is white and the background is black. You may prefer to invert these images when you display or print them, or else display them using a colormap. See Chapter 2 for more information about displaying binary images. The remainder of this chapter describes the functions in the Image Processing Toolbox that perform various types of binary image operations. These operations include: • Morphological operations • Object-based operations • Feature measurement • Lookup table operations www.MatlabSite.com 8-3 8 Binary Image O perations M orphologica l O pera tions Morphological operations are methods for processing binary images based on shapes. These operations take a binary image as input, and return a binary image as output. The value of each pixel in the output image is based on the corresponding input pixel and its neighbors. By choosing the neighborhood shape appropriately, you can construct a morphological operation that is sensitive to specific shapes in the input image. Dila tion a nd Erosion The main morphological operations are dilation and erosion. Dilation and erosion are related operations, although they produce very different results. Dilation adds pixels to the boundaries of objects (i.e., changes them from off to on), while erosion removes pixels on object boundaries (changes them from on to off). Each dilation or erosion operation uses a specified neighborhood. The state of any given pixel in the output image is determined by applying a rule to the neighborhood of the corresponding pixel in the input image. The rule used defines the operation as a dilation or an erosion: • For dilation, if any pixel in the input pixel’s neighborhood is on, the output pixel is on. Otherwise, the output pixel is off. • For erosion, if every pixel in the input pixel’s neighborhood is on, the output pixel is on. Otherwise, the output pixel is off. The neighborhood for a dilation or erosion operation can be of arbitrary shape and size. The neighborhood is represented by a structuring element, which is a matrix consisting of only 0’s and 1’s. The center pixel in the structuring element represents the pixel of interest, while the elements in the matrix that are on (i.e., = 1) define the neighborhood. The center pixel is defined as floor((size(SE)+1)/2), where SE is the structuring element. For example, in a 4-by-7 structuring element, the center pixel is (2,4). When you construct the structuring element, you should make sure that the pixel of interest is actually the center pixel. You can do this by adding rows or columns of 0’s, if necessary. For example, suppose you want the neighborhood to consist of a 3-by-3 block of pixels, with the pixel of interest in the upper-left corner of the block. The structuring element would not be 8-4 www.MatlabSite.com M orphological O perations ones(3), because this matrix has the wrong center pixel. Rather, you could use this matrix as the structuring element: 0 0 0 0 0 1 1 1 0 1 1 1 0 1 1 1 For erosion, the neighborhood consists of the on pixels in the structuring element. For dilation, the neighborhood consists of the on pixels in the structuring element rotated 180 degrees. (The center pixel is still selected before the rotation.) Suppose you want to perform an erosion operation. The figure below shows a sample neighborhood you might use. Each neighborhood pixel is indicated by an x, and the center pixel is the one with a circle: x x x x x The structuring element is therefore: 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 The state (i.e., on or off) of any given pixel in the output image is determined by applying the erosion rule to the neighborhood pixels for the corresponding pixel in the input image. For example, to determine the state of the pixel (4,6) in the output image: www.MatlabSite.com 8-5 8 Binary Image O perations • Overlay the structuring element on the input image, with the center pixel of the structuring element covering the pixel (4,6). • Look at the pixels in the neighborhood of the input pixel. These are the five pixels covered by 1’s in the structuring element. In this case the pixels are: (2,4), (3,5), (4,6), (5,7), (6,8). If all of these pixels are on, then set the pixel in the output image (4,6) to on. If any of these pixels is off, then set the pixel (4,6) in the output image to off. You perform this procedure for each pixel in the input image to determine the state of each corresponding pixel in the output image. Note that for pixels on borders of the image, some of the 1’s in the structuring element are actually outside the image. These elements are assumed to cover off pixels. As a result, the output image will usually have a black border, as in the example below. The Image Processing Toolbox performs dilation through the dilate function, and erosion through the erode function. Each of these functions takes an input image and a structuring element as input, and returns an output image. This example illustrates the erosion operation described above: BW1 = imread('circbw.tif'); SE = eye(5); BW2 = erode(BW1,SE); imshow(BW1) figure, imshow(BW2) Notice the diagonal streaks in the output image (on the right). These are due to the shape of the structuring element. 8-6 www.MatlabSite.com M orphological O perations Rela ted O pera tions There are many other types of morphological operations in addition to dilation and erosion. However, many of these operations are just modified dilations or erosions, or combinations of dilation and erosion. For example, closure consists of a dilation operation followed by erosion with the same structuring element. A related operation, opening, is the reverse of closure; it consists of erosion followed by dilation. For example, suppose you want to remove all the circuit lines from the original circuit image, leaving only the rectangular outlines of microchips. You can accomplish this through opening. To perform the opening, you begin by choosing a structuring element. This structuring element should be large enough to remove the lines when you erode the image, but not large enough to remove the rectangles. It should consist of all 1’s, so it removes everything but large continuous patches of foreground pixels. Therefore, you create the structuring element like this: SE = ones(40,30); Next, you perform the erosion. This removes all of the lines, but also shrinks the rectangles: BW2 = erode(BW1,SE); imshow(BW2) www.MatlabSite.com 8-7 8 Binary Image O perations Finally, you perform dilation, using the same structuring element, to restore the rectangles to their original sizes: BW3 = dilate(BW2,SE); imshow(BW3) Predefined O perations You can use dilate and erode to implement any morphological operation that can be defined as a set of dilations and erosions. However, there are certain operations that are so common that the toolbox provides them as predefined procedures. These operations are available through the bwmorph function. bwmorph provides eighteen predefined operations, including opening and closure. 8-8 www.MatlabSite.com M orphological O perations For example, suppose you want to reduce all objects in the circuit image to lines, without changing the essential structure (topology) of the image. This process is known as skeletonization. You can use bwmorph to do this: BW1 = imread('circbw.tif'); BW2 = bwmorph(BW1,'skel',Inf); imshow(BW1) figure, imshow(BW2) The third argument to bwmorph indicates the number of times to perform the operation. For example, if this value is 2, the operation is performed twice, with the result of the first operation being used as the input for the second operation. In the example above, the value is Inf. In this case bwmorph performs the operation repeatedly until it no longer changes. For more information about the predefined operations available, see the reference entry for bwmorph in Chapter 11. www.MatlabSite.com 8-9 8 Binary Image O perations O bject-Ba sed O pera tions In a binary image, an object is any set of connected on pixels. For example, this matrix represents a binary image containing a single object, a 3-by-3 square. The rest of the image is background: 0 0 0 0 0 0 0 0 1 1 1 0 0 0 1 1 1 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 This section discusses the types of neighborhoods used for object-based operations, and describes how to use toolbox functions to perform: • Perimeter determination • Binary flood fill • Connected-components labeling • Object selection 4 - a nd 8 -Connected N eighborhoods For many operations, distinguishing objects depends on the convention used to decide whether pixels are connected. There are two different conventions typically used: 4-connected or 8-connected neighborhoods. In an 8-connected neighborhood, all of the pixels that touch the pixel of interest are considered, including those on the diagonals. This means that if two adjoining pixels are on, they are part of the same object, regardless of whether they are connected along the horizontal, vertical, or diagonal direction. 8-10 www.MatlabSite.com O bject-Based O perations This figure illustrates an 8-connected neighborhood: r–1, c–1 r–1, c r–1, c+1 r, c–1 r, c r, c+1 r+1, c–1 r+1, c r+1, c+1 In a 4-connected neighborhood, the pixels along the diagonals are not considered. This means that two adjoining pixels are part of the same object only if they are both on and are connected along the horizontal or vertical direction. This figure illustrates a 4-connected neighborhood: r–1, c r, c–1 r, c r, c+1 r+1, c www.MatlabSite.com 8-11 8 Binary Image O perations The type of neighborhood you choose affects the number of objects there are in an image and the boundaries of those objects. Therefore, the results of the object-based operations often differ for the two types of neighborhoods. For example, this matrix represents a binary image that has one 8-connected object or two 4-connected objects: 0 0 0 0 0 0 0 1 1 0 0 0 0 1 1 0 0 0 0 0 0 1 1 0 0 0 0 1 1 0 0 0 0 0 0 0 Perim eter Determ ina tion The bwperim function determines the perimeter pixels of the objects in a binary image. You can use either a 4- or 8-connected neighborhood for perimeter determination. A pixel is considered a perimeter pixel if it satisfies both of these criteria: • It is an on pixel. • One (or more) of the pixels in its neighborhood is off. This example finds the perimeter pixels in the circuit image: BW1 = imread('circbw.tif'); BW2 = bwperim(BW1); imshow(BW1) figure, imshow(BW2) 8-12 www.MatlabSite.com O bject-Based O perations Flood Fill The bwfill function performs a flood-fill operation on a binary image. You specify a background pixel as a starting point, and bwfill changes connected background pixels (0’s) to foreground pixels (1’s), stopping when it reaches object boundaries. The boundaries are determined based on the type of neighborhood you specify. This operation can be useful in removing irrelevant artifacts from images. For example, suppose you have a binary image, derived from a photograph, in which the foreground objects represent spheres. In the binary image, these objects should appear as circles, but instead are donut shaped because of reflections in the original photograph. Before doing any further processing of the image, you may want to first fill in the “donut holes” using bwfill. bwfill differs from the other object-based operations in that it operates on background pixels, rather than the foreground. If the foreground is 8-connected, the background is 4-connected, and vice versa. Note, however, that as with the other object-based functions, you specify the connectedness of the foreground when you call bwfill. The implications of 4- vs. 8-connected foreground can be illustrated with this matrix: BW1 = 0 0 0 0 0 0 0 0 0 1 1 1 1 1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Regardless of whether the foreground is 4-connected or 8-connected, this image contains a single object. However, the topology of the object differs depending on the type of neighborhood. If the foreground is 8-connected, the object is a closed contour, and there are two separate background elements (the part inside the loop and the part outside). If the foreground is 4-connected, the contour is open, and there is only one background element. www.MatlabSite.com 8-13 8 Binary Image O perations Suppose you call bwfill, specifying the pixel BW1(4,3) as the starting point: bwfill(BW1,4,3) ans = 0 0 0 0 0 0 0 0 0 1 1 1 1 1 0 0 0 1 1 1 1 1 0 0 0 1 1 1 1 1 0 0 0 1 1 1 1 1 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 bwfill fills in just the inside of the loop, because bwfill uses an 8-connected foreground by default. If you specify a 4-connected foreground instead, bwfill fills in the entire image, because the entire background is a single 8-connected element: bwfill(BW1,4,3,4) ans = 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Note that unlike other binary image operations, bwfill pads with 1’s rather than 0’s along the image border. This prevents the fill operation from wrapping around the border. You can also use bwfill interactively, selecting starting pixels with a mouse. See the reference entry for bwfill in Chapter 11 for more information about this function. 8-14 www.MatlabSite.com O bject-Based O perations Connected-Com ponents La beling You can use the bwlabel function to perform connected-components labeling, which is a method for indicating each discrete object in a binary image. You specify the input binary image and the type of neighborhood, and bwlabel returns a matrix of the same size as the input image. The different objects in the input image are distinguished by different integer values in the output matrix. For example, suppose you have this binary image: BW1 = 0 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 0 1 1 1 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 0 1 1 1 0 0 1 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 1 1 0 0 0 0 0 You call bwlabel, specifying 4-connected neighborhoods: bwlabel(BW1,4) ans = 0 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 0 1 1 1 0 0 0 0 0 0 0 0 2 2 2 0 0 0 0 0 2 2 2 0 0 3 0 0 0 0 0 0 0 3 3 0 0 0 0 0 0 3 3 0 0 0 0 0 In the output matrix, the 1’s represent one object, the 2’s a second object, and the 3’s a third. Notice that if you had used 8-connected neighborhoods (the default), there would be only two objects, because the first and second objects would be a single object, connected along the diagonal. www.MatlabSite.com 8-15 8 Binary Image O perations The output matrix is not a binary image, and its class is double. A useful approach to viewing it is to display it as an pseudocolor indexed image, first adding 1 to each element, so that the data is in the proper range. Each object displays in a different color, so the objects are easier to distinguish than in the original image. The example below illustrates this technique. Notice that this example uses a colormap in which the first color (the background) is black and the other colors are easily distinguished: X = bwlabel(BW1,4); map = [0 0 0;jet(3)]; imshow(X+1,map,'notruesize') O bject Selection You can use the bwselect function to select individual objects in a binary image. You specify pixels in the input image, and bwselect returns a binary image that includes only those objects from the input image that contain one of the specified pixels. You can specify the pixels either noninteractively or with a mouse. For example, suppose you want to select 8-connected objects in the image displayed in the current axes. You type: BW2 = bwselect(8); The cursor changes to a cross-hair when it is over the image. Click on the objects you want to select; bwselect displays a small star over each pixel you click on. When you are done, press Return. bwselect returns a binary image consisting of the objects you selected, and removes the stars. 8-16 www.MatlabSite.com O bject-Based O perations See the reference entry for bwselect in Chapter 11 for more information about this function. www.MatlabSite.com 8-17 8 Binary Image O perations Fea ture M ea surement When you process an image, you may want to obtain information about how certain features of the image change. For example, when you perform dilation, you may want to determine how many pixels are changed from off to on by the operation, or to see if the number of objects in the image changes. This section describes two functions, bwarea and bweuler, that return two common measures of binary images: the image area and the Euler number. In addition, you can use the imfeature function, in combination with bwlabel, to compute measurements of various features in a binary image. See the reference entry for imfeature in Chapter 11 for more information. Im a ge Area The bwarea function returns the area of a binary image. The area is a measure of the size of the foreground of the image. Roughly speaking, the area is the number of on pixels in the image. bwarea does not simply count the number of on pixels, however. Rather, bwarea weights different pixel patterns unequally when computing the area. This weighting compensates for the distortion that is inherent in representing a continuous image with discrete pixels. For example, a diagonal line of 50 pixels is longer than a horizontal line of 50 pixels. As a result of the weighting bwarea uses, the horizontal line has area of 50, but the diagonal line has area of 62.5. This example uses bwarea to determine how much the area of the circuit image increases after a dilation operation: BW1 = imread('circbw.tif'); SE = ones(5); BW2 = dilate(BW1,SE); increase = (bwarea(BW2) – bwarea(BW1))/bwarea(BW1); increase = 0.3456 The dilation increases the area by about 35 percent. For more information about the weighting pattern used by bwarea, see the reference entry for bwarea in Chapter 11. 8-18 www.MatlabSite.com Feature M easurement Euler N um ber The bweuler function returns the Euler number for a binary image. The Euler number is a measure of the topology of an image. It is defined as the total number of objects in the image minus the number of holes in those objects. You can use either 4- or 8-connected neighborhoods. This example computes the Euler number for the circuit image, using 8-connected neighborhoods: BW1 = imread('circbw.tif'); eul = bweuler(BW1,8) eul = –85 In this example, the Euler number is negative, indicating that the number of holes is greater than the number of objects. www.MatlabSite.com 8-19 8 Binary Image O perations Look up-Ta ble O pera tions Certain binary image operations can be implemented most easily through lookup tables. A lookup table is a column vector in which each element represents the value to return for one possible combination of pixels in a neighborhood. You can use the makelut function to create lookup tables for various operations. makelut creates lookup tables for 2-by-2 and 3-by-3 neighborhoods. This figure illustrates these types of neighborhoods. Each neighborhood pixel is indicated by an x, and the center pixel is the one with a circle: x x x x x x x x x x x x x 2 -by-2 neighborhood 3 -by-3 neighborhood For a 2-by-2 neighborhood, there are 16 possible permutations of the pixels in the neighborhood. Therefore, the lookup table for this operation is a 16-element vector. For a 3-by-3 neighborhood, there are 512 permutations, so the lookup table is a 512-element vector. Once you create a lookup table, you can use it to perform the desired operation by using the applylut function. The example below illustrates using lookup-table operations to modify an image containing text. You begin by writing a function that returns 1 if three or more pixels in the 3-by-3 neighborhood are 1, and 0 otherwise. You then call makelut, passing in this function as the first argument, and using the second argument to specify a 3-by-3 lookup table: f = inline('sum(x(:)) >= 3'); lut = makelut(f,3); lut is returned as a 512-element vector of 1’s and 0’s. Each value is the output from the function for one of the 512 possible permutations. 8-20 www.MatlabSite.com Lookup-Table O perations You then perform the operation using applylut: BW1 = imread('text.tif'); BW2 = applylut(BW1,lut); imshow(BW1) figure, imshow(BW2) For information about how applylut maps pixel combinations in the image to entries in the lookup table, see the reference entry for applylut in Chapter 11. Note that you cannot use makelut and applylut for neighborhoods of sizes other than 2-by-2 or 3-by-3. These functions support only 2-by-2 and 3-by-3 neighborhoods, because lookup tables are not practical for neighborhoods larger than 3-by-3. For example, a lookup table for a 4-by-4 neighborhood would have 65,536 entries. www.MatlabSite.com 8-21 8 Binary Image O perations 8-22 www.MatlabSite.com 9 Region-Based Processing Overview . . . . . . . . . . . . . . . . . . . . . 9-2 Specifying a Region of Interest . . . . . . . . . . . 9-3 Selecting a Polygon . . . . . . . . . . . . . . . . . . 9-3 Other Selection Methods . . . . . . . . . . . . . . . 9-5 Filtering a Region Filling a Region . . . . . . . . . . . . . . . . . 9-6 . . . . . . . . . . . . . . . . . . 9-8 www.MatlabSite.com 9 Region-Based Processing O verview This chapter describes operations that you can perform on a selected region of an image. It discusses these topics: • Specifying a region • Filtering a region • Filling a region For an interactive demonstration of region-based processing, try running roidemo. 9-2 www.MatlabSite.com Specifying a Region of Interest Specifying a Region of Interest A region of interest is a portion of an image that you want to filter or perform some other operation on. You define a region of interest by creating a binary mask, which is a binary image of the same size as the image you want to process. The mask contains 1’s for all pixels that are part of the region of interest, and 0’s everywhere else. This section discusses methods for creating binary masks. Selecting a Poly gon You can use the roipoly function to specify a polygonal region of interest. If you call roipoly with no input arguments, the cursor changes to a cross hair when it is over the image displayed in the current axes. You can then specify the vertices of the polygon by clicking on points in the image with the mouse. When you are done selecting vertices, press Return; roipoly returns a binary image, of the same size as the input image, containing 1’s inside the specified polygon, and 0’s everywhere else. www.MatlabSite.com 9-3 9 Region-Based Processing The example below illustrates using roipoly to create a binary mask. The border of the selected region is shown in red on the original image. I = imread('pout.tif'); imshow(I) BW = roipoly; imshow(BW) You can also use roipoly noninteractively. See the reference entry for roipoly in Chapter 11 for more information. 9-4 www.MatlabSite.com Specifying a Region of Interest O ther Selection M ethods roipoly provides an easy way to create a binary mask. However, you can use any binary image as a mask, provided that the binary image is the same size as the image being filtered. For example, suppose you want to filter the intensity image I, but only filter those pixels whose values are greater than 0.5. You could create the appropriate mask with this command: BW = (I > 0.5); You can also use the roicolor function to define the region of interest based on a color or intensity range. For more information, see the reference entry for this function in Chapter 11. www.MatlabSite.com 9-5 9 Region-Based Processing Filtering a Region You can use the roifilt2 function to process a region of interest. When you call roifilt2, you specify an intensity image, a binary mask, and a filter. roifilt2 filters the input image, and returns an image that consists of filtered values for pixels where the binary mask contains 1’s, and unfiltered values for pixels where the binary mask contains 0’s. This type of operation is called masked filtering. This example uses the mask created on page 9-4 to increase the contrast of the logo on the girl’s coat. h = fspecial('unsharp'); I2 = roifilt2(h,I,BW); imshow(I) figure, imshow(I2) roifilt2 also has a form that you can use to specify a function to operate on the region of interest. In the example below, the imadjust function is used to 9-6 www.MatlabSite.com Filtering a Region lighten parts of an image. The mask in the example is a binary image containing text. The resulting image has the text imprinted on it. BW = imread('text.tif'); I = imread('cameraman.tif'); f = inline('imadjust(x,[],[],0.3)'); I2 = roifilt2(I,BW,f); imshow(I2) Note that roifilt2 is best suited to operations that return data in the same range as in the original image, because the output image takes some of its data directly from the input image. Certain filtering operations can result in values outside the normal image data range (i.e., [0,1] for images of class double, [0,255] for images of class uint8). For more information about roifilt2, see the reference entry in Chapter 11. www.MatlabSite.com 9-7 9 Region-Based Processing Filling a Region You can use the roifill function to fill a region of interest, interpolating from the borders of the region. This function is useful for image editing, to remove extraneous details or artifacts. roifill performs the fill using an interpolation method based on Laplace’s equation. This method results in the smoothest possible fill, given the values on the boundary of the region. As with roipoly, you select the region of interest with the mouse. When you complete the selection, roifill returns an image with the selected region filled in. 9-8 www.MatlabSite.com Filling a Region This example uses roifill to modify the trees image. The border of the selected region is shown in red on the original image. load trees I = ind2gray(X,map); imshow(I) I2 = roifill; imshow(I2) www.MatlabSite.com 9-9 9 Region-Based Processing 9-10 www.MatlabSite.com 10 Color Overview . . . . . . . . . . . . . . . . . . . . . 10-2 . . . . . . . . 10-3 Working with Different Color Depths Reducing the Number of Colors in an Image Using rgb2ind . . . . . . . . . . . . . . . Using imapprox . . . . . . . . . . . . . . Dithering . . . . . . . . . . . . . . . . . Converting to Other Color Spaces NTSC Format . . . . . . . . . . YCbCr Format . . . . . . . . . . HSV Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-9 . 10-9 10-10 10-10 www.MatlabSite.com 10-5 10-5 10-7 10-8 10 C olor O verview This chapter describes toolbox functions that help you work with color image data. It discusses these topics: • Working on systems with different color depths • Reducing the number of colors in an image • Converting an image to a different color space Note that “color” includes shades of gray, so much of the discussion in this chapter applies to grayscale images as well as color images. For additional information about how MATLAB handles color, see the Using MATLAB Graphics manual. 10-2 www.MatlabSite.com W orking w ith Different C olor Depths W ork ing w ith Different Color Depths Most computer displays use either 8, 16, or 24 bits per screen pixel. The number of bits per pixel determines how many different colors can be displayed. 24-bit displays provide optimal color resolution. A 24-bit display uses 8 bits for each of the three color components; therefore there are 256 levels each of red, green, and blue. This makes it possible to support 16,777,216 (i.e., 224) distinct colors. Of these colors, 256 are shades of gray. (These are the colors where R=G=B.) 16-bit displays also provide a large number of colors. A 16-bit display uses 5 bits for each color component, resulting in 32 shades each of red, green, and blue. (Some 16-bit displays use the remaining bit to increase the number of shades of green to 64.) 16-bit displays support 32,768 (i.e., 215) distinct colors (or twice this number if all 16 bits are used). Of these colors, 32 are shades of gray. 8-bit displays support a much more limited number of colors. An 8-bit display can produce any of the colors available on a 24-bit display, but only 256 distinct colors can appear at one time. There are 256 shades of gray available, but using all of them takes up all of the available color slots. To determine your system’s color depth, enter this MATLAB command: get(0,'ScreenDepth') MATLAB returns an integer representing the number of bits per pixel. Note that on some systems, MATLAB may return 32. These systems produce 24-bit color; the remaining 8 bits are for an alpha (transparency) channel that MATLAB does not currently use. Depending on your system, you may be able to choose the color depth you want to use. (There may be tradeoffs between color depth and screen resolution.) In general, 24-bit display mode produces the best results. If you need to use a lower color depth, 16-bit is generally preferable to 8-bit. However, keep in mind that 16-bit color has certain limitations: www.MatlabSite.com 10-3 10 C olor • An image may have finer gradations of color than a 16-bit display can represent. If a color is unavailable, MATLAB uses the closest approximation. • There are only 32 shades of gray available. If you are working primarily with grayscale images, you may get better display results using 8-bit display mode. Regardless of the number of colors your system can display, MATLAB can store and process images with up to 224 colors (or even more, if the image data is of class double). These images display best on systems with 24-bit color, and usually look fine on 16-bit systems as well. (On a 16-bit system, if an image uses colors that the system is not capable of displaying, MATLAB uses color approximation to display the image.) If your display supports only 8-bit color, you may run into problems due to the limited number of color slots. See Chapter 2 for information about working around these problem. Also, see the next section for information about reducing the number of colors used by an image. 10-4 www.MatlabSite.com Reducing the N umber of C olors in an Image Reducing the N umber of Colors in a n Ima ge On systems with 24-bit color, RGB (truecolor) images can display up to 224 (i.e., 16,777,216) colors. On systems with lower color depths, RGB images display reasonably well, because MATLAB automatically uses color approximation and dithering if needed. MATLAB can also read and write these images using imread and imwrite, without any loss of color resolution. Indexed images, however, may cause problems if they have a large number of colors. In general, you should limit indexed images to 256 colors, because: • On systems with 8-bit color, indexed images with more than 256 colors do not display properly. • On some platforms, colormaps cannot exceed 256 entries. • If an indexed image has more than 256 colors, MATLAB cannot store the image data in a uint8 array, and must use an array of class double instead. • Most image file formats limit indexed images to 256 colors. imwrite produces an error if you try to write an indexed image with more colors than this to one of these formats. This section describes how to reduce the number of colors in an indexed or RGB image: • The rgb2ind function converts an RGB image to an indexed image, and provides a number of methods for approximating the original image with fewer colors. • The imapprox function takes an indexed image and returns a new indexed image that uses fewer colors, using the same color-approximation methods that rgb2ind uses. In addition, this section includes a discussion of dithering, which is used by these functions to increase the apparent color resolution of an image. Using r gb2 ind rgb2ind converts an RGB image to an indexed image, reducing the number of colors in the process. This function provides these methods for approximating the colors in the original image: www.MatlabSite.com 10-5 10 C olor • Quantization • Colormap mapping The quality of the resulting image depends on the approximation method you use, the range of colors in the input image, and whether or not you use dithering. Note that different methods work better for different images. See page 10-8 for information about dithering. Q uantization In general, the most effective approach to reducing the number of colors in an image is quantization. This approach involves cutting the RGB color cube into a number of smaller regions, and then mapping all colors that fall within a given region to the value at the center of that region. Quantization ensures that the colors in the output image do not deviate too much from those in the input image, and also enables you to control the maximum number of output colors. rgb2ind supports two quantization methods: uniform quantization and minimum variance quantization. These methods differ in the approach used to divide up the RGB cube. With uniform quantization, the color cube is cut up into equal-sized cubes. With minimum variance quantization, the color cube is cut up into smaller boxes (not necessarily cubes) of different sizes; the sizes of the boxes depend on how the colors are distributed in the image. For a given number of colors, minimum variance quantization produces better results than uniform quantization, because it takes into account the actual data. Minimum variance quantization allocates more of the colormap entries to colors that appear frequently in the input image, and fewer to colors that appear infrequently. For example, if the input image has many shades of green and few shades of red, there will be many more greens than reds in the output colormap. The resulting color resolution is higher than with uniform quantization. The computation takes longer, however. To perform uniform quantization, you call rgb2ind and specify a tolerance. The tolerance determines the size of the cubes that the RGB color cube is cut into. For example, if you specify a tolerance of 0.1, the edges of the cubes are one-tenth the length of the RGB cube. The total number of small cubes is: n = (floor(1/tol)+1)^3 Each cube represents a single color in the output image. Therefore, the maximum length of the colormap is n. rgb2ind removes any colors that do not appear in the input image, so the actual colormap may be much smaller than n. 10-6 www.MatlabSite.com Reducing the N umber of C olors in an Image To perform minimum variance quantization, you call rgb2ind and specify the maximum number of colors in the output image’s colormap. The number you specify determines how many cubes the RGB color cube is divided into. The resulting colormap usually has the number of entries you specify, because the color cube is divided so that each region contains at least one color that appears in the input image. If the input image actually uses fewer colors than the number you specify, the output colormap is also smaller. In this case, the output image has all of the colors of the input image, and there is no loss of color resolution in the conversion. These commands use minimum variance quantization to create an indexed image with 185 colors: RGB = imread('flowers.tif'); [X,map] = rgb2ind(RGB,185); Colormap M apping If you specify the actual colormap to use, rgb2ind uses colormap mapping to find the colors in the specified colormap that best match the colors in the RGB image. This method is useful if you need to create images that use a fixed colormap. For example, if you want to display multiple indexed images on an 8-bit display, you can avoid color problems by having them all use the same colormap. Colormap mapping produces a good approximation if the specified colormap has similar colors to those in the RGB image. If the colormap does not have similar colors to those in the RGB image, this method produces poor results. This example illustrates mapping two images to the same colormap. The colormap specified includes colors all throughout the RGB cube, so the output images are able to reasonably approximate the input images. RGB1 RGB2 X1 = X2 = = imread('autumn.tif'); = imread('flowers.tif'); rgb2ind(RGB1,colorcube(128)); rgb2ind(RGB2,colorcube(128)); Using im a pprox imapprox is based on rgb2ind, and provides the same approximation methods. Use imapprox when you need to reduce the number of colors in an indexed image. Essentially, imapprox first calls ind2rgb to convert the image to RGB www.MatlabSite.com 10-7 10 C olor format, and then calls rgb2ind to return a new indexed image with fewer colors. For example, these commands create a version of the trees image with 64 colors, rather than the original 128: load trees [Y,newmap] = imapprox(X,map,64); Dithering When you use rgb2ind or imapprox to reduce the number of colors in an image, the resulting image may look inferior to the original, because some of the colors are lost. rgb2ind and imapprox both perform dithering to increase the apparent color resolution of the output image. Dithering changes the colors of pixels in a neighborhood, so that the average color around each pixel approximates the original RGB color. Dithering provides this extra color resolution at the expense of spatial resolution, and it adds high frequency noise, but it typically results in a more visually pleasing approximation. If you do not want the output image to be dithered, you can specify this when you call rgb2ind or imapprox. For example: [X,map] = rgb2ind(RGB,128,'nodither'); 10-8 www.MatlabSite.com C onverting to O ther C olor Spaces Converting to O ther Color Spa ces The Image Processing Toolbox represents colors as RGB values, either directly (in an RGB image) or indirectly (in an indexed image). However, there are other models besides RGB for representing colors numerically. For example, a color can be represented by its hue, saturation, and value components (HSV) instead. The various models for color data are called color spaces. The functions in the Image Processing Toolbox that work with color assume that images use the RGB color space. The toolbox provides support for other color spaces though a set of conversion functions. You can use these functions to convert between RGB and these color spaces: • National Television Systems Committee (NTSC) format • YCbCr format • Hue, saturation, value (HSV) format This section describes these color spaces and the conversion routines for working with them. N TSC Form a t The NTSC format is used in televisions in the United States. One of the main advantages of this format is that grayscale information is separated from color data, so the same signal can be used for both color and black and white sets. In this NTSC format, image data consists of three components: luminance (Y), hue (I), and saturation (Q). The first component, luminance, represents grayscale information, while the last two components make up chrominance (color information). The function rgb2ntsc converts colormaps or RGB images to the NTSC color space. ntsc2rgb performs the reverse operation. For example, these commands convert the flowers image to NTSC format: RGB = imread('flowers.tif'); YIQ = rgb2ntsc(RGB); Because luminance is one of the components of the NTSC format, the RGB to NTSC conversion is also useful for isolating the gray level information in an image. The toolbox functions rgb2gray and ind2gray use the rgb2ntsc function to extract the grayscale information from a color image. www.MatlabSite.com 10-9 10 C olor For example, these commands are equivalent to calling rgb2gray: YIQ = rgb2ntsc(RGB); I = YIQ(:,:,1); YCbCr Form a t The YCbCr format is widely used for digital video. In this format, luminance information is stored as a single component (Y), and chrominance information is stored as two color difference components (Cb and Cr). Cb represents the difference between the blue component and a reference value, and Cr represents the difference between the red component and a reference value. YCbCr data can be double precision, but the color space is particularly well suited to uint8 data. For uint8 images, the data range for Y is [16,235], and the range for Cb and Cr is [16,240]. YCbCr leaves room at the top and bottom of the full uint8 range so that additional (non-image) information can be included in a video stream. The function rgb2ycbcr converts colormaps or RGB images to the YCbCr color space. ycbcr2rgb performs the reverse operation. For example, these commands convert the flowers image to YCbCr format: RGB = imread('flowers.tif'); YCBCR = rgb2ycbcr(RGB); H SV Form a t The functions rgb2hsv and hsv2rgb convert images between the RGB and hue, saturation, value (HSV) color spaces. The HSV color space is often used for picking colors (e.g., of paints or inks) from a color wheel or palette, because it corresponds better to how people experience color than the RGB color space does. As hue varies from 0 to 1.0, the corresponding colors vary from red, through yellow, green, cyan, blue, and magenta, back to red. As saturation varies from 0 to 1.0, the corresponding colors vary from unsaturated (shades of gray) to fully saturated (no white component). As value, or brightness, varies from 0 to 1.0, the corresponding colors become increasingly brighter. 10-10 www.MatlabSite.com C onverting to O ther C olor Spaces The figure below illustrates the HSV color space: Hue Hue 1 0 1 Saturation Value 0 1 .5 0.5 0 0 These commands convert an RGB image to HSV color space: RGB = imread('flowers.tif'); HSV = rgb2hsv(RGB); www.MatlabSite.com 10-11 10 C olor 10-12 www.MatlabSite.com 11 Reference www.MatlabSite.com 11 Reference This chapter provides detailed descriptions of the functions in the Image Processing Toolbox. It begins with a list of functions grouped by subject area and continues with the reference entries in alphabetical order. The list of functions includes all functions in the Image Processing Toolbox itself, plus a few functions in MATLAB that are especially useful for image processing. All of the functions listed have reference entries, with these exceptions: • The MATLAB functions image, imagesc, and interp2. For information about these functions, see the online MATLAB Function Reference. • The Image Processing Toolbox demo functions and slideshow functions. For information about any of these functions, run the demo or slideshow and click the Info button. Im a ge Displa y 11-2 colorbar Display colorbar (MATLAB) getimage Get image data from axes image Create and display image object (MATLAB) imagesc Scale data and display as image (MATLAB) immovie Make movie from multiframe indexed image imshow Display image montage Display multiple image frames as rectangular montage subimage Display multiple images in single figure truesize Adjust display size of image warp Display image as texture-mapped surface zoom Zoom in and out of image or 2-D plot (MATLAB) www.MatlabSite.com Im a ge File I/ O imfinfo Return information about image file (MATLAB) imread Read image file (MATLAB) imwrite Write image file (MATLAB) Geom etric O pera tions imcrop Crop image imresize Resize image imrotate Rotate image interp2 2-D data interpolation (MATLAB) Pix el Va lues a nd Sta tistics corr2 Compute 2-D correlation coefficient imcontour Create contour plot of image data imfeature Compute feature measurements for image regions imhist Display histogram of image data impixel Determine pixel color values improfile Compute pixel-value cross-sections along line segments mean2 Compute mean of matrix elements pixval Display information about image pixels std2 Compute standard deviation of matrix elements www.MatlabSite.com 11-3 11 Reference Im a ge Ana lysis edge Find edges in intensity image qtdecomp Perform quadtree decomposition qtgetblk Get block values in quadtree decomposition qtsetblk Set block values in quadtree decomposition Im a ge Enha ncement histeq Enhance contrast using histogram equalization imadjust Adjust image intensity values or colormap imnoise Add noise to an image medfilt2 Perform 2-D median filtering ordfilt2 Perform 2-D order-statistic filtering wiener2 Perform 2-D adaptive noise-removal filtering Linea r Filtering 11-4 conv2 Perform 2-D convolution (MATLAB) convmtx2 Compute 2-D convolution matrix convn Perform N-D convolution (MATLAB) filter2 Perform 2-D filtering (MATLAB) fspecial Create predefined filters www.MatlabSite.com Linea r 2 -D Filter Design freqspace Determine 2-D frequency response spacing (MATLAB) freqz2 Compute 2-D frequency response fsamp2 Design 2-D FIR filter using frequency sampling ftrans2 Design 2-D FIR filter using frequency transformation fwind1 Design 2-D FIR filter using 1-D window method fwind2 Design 2-D FIR filter using 2-D window method Im a ge Tra nsforms dct2 Compute 2-D discrete cosine transform dctmtx Compute discrete cosine transform matrix fft2 Compute 2-D fast Fourier transform (MATLAB) fftn Compute N-D fast Fourier transform (MATLAB) fftshift Reverse quadrants of output of FFT (MATLAB) idct2 Compute 2-D inverse discrete cosine transform ifft2 Compute 2-D inverse fast Fourier transform (MATLAB) ifftn Compute N-D inverse fast Fourier transform (MATLAB) iradon Compute inverse Radon transform phantom Generate a head phantom image radon Compute Radon transform www.MatlabSite.com 11-5 11 Reference N eighborhood a nd Block Processing bestblk Choose block size for block processing blkproc Implement distinct block processing for image col2im Rearrange matrix columns into blocks colfilt Perform neighborhood operations using columnwise functions im2col Rearrange image blocks into columns nlfilter Perform general sliding-neighborhood operations Bina ry Im a ge O pera tions 11-6 applylut Perform neighborhood operations using lookup tables bwarea Compute area of objects in binary image bweuler Compute Euler number of binary image bwfill Fill background regions in binary image bwlabel Label connected components in binary image bwmorph Perform morphological operations on binary image bwperim Determine perimeter of objects in binary image bwselect Select objects in binary image dilate Perform dilation on binary image erode Perform erosion on binary image makelut Construct lookup table for use with applylut www.MatlabSite.com Region-Ba sed Processing roicolor Select region of interest, based on color roifill Smoothly interpolate within arbitrary region roifilt2 Filter a region of interest roipoly Select polygonal region of interest Colorm a p M a nipula tion brighten Brighten or darken colormap (MATLAB) cmpermute Rearrange colors in colormap cmunique Find unique colormap colors and corresponding image colormap Set or get color lookup table (MATLAB) imapprox Approximate indexed image by one with fewer colors rgbplot Plot RGB colormap components (MATLAB) Color Spa ce Conversions hsv2rgb Convert HSV values to RGB color space (MATLAB) ntsc2rgb Convert NTSC values to RGB color space rgb2hsv Convert RGB values to HSV color space (MATLAB) rgb2ntsc Convert RGB values to NTSC color space rgb2ycbcr Convert RGB values to YCbCr color space ycbcr2rgb Convert YCbCr values to RGB color space www.MatlabSite.com 11-7 11 Reference Im a ge Types a nd Type Conversions dither Convert image using dithering gray2ind Convert intensity image to indexed image grayslice Create indexed image from intensity image by thresholding im2bw Convert image to binary image by thresholding im2double Convert image array to double precision im2uint8 Convert image array to 8-bit unsigned integers ind2gray Convert indexed image to intensity image ind2rgb Convert indexed image to RGB image isbw Return true for binary image isgray Return true for intensity image isind Return true for indexed image isrgb Return true for RGB image mat2gray Convert matrix to intensity image rgb2gray Convert RGB image or colormap to grayscale rgb2ind Convert RGB image to indexed image Toolbox Preferences 11-8 iptgetpref Get value of Image Processing Toolbox preference iptsetpref Set value of Image Processing Toolbox preference www.MatlabSite.com Dem os dctdemo 2-D DCT image compression demo edgedemo Edge detection demo firdemo 2-D FIR filtering and filter design demo imadjdemo Intensity adjustment and histogram equalization demo nrfiltdemo Noise reduction filtering demo qtdemo Quadtree decomposition demo roidemo Region-of-interest processing demo Slide Show s ipss001 Region labeling of steel grains ipss002 Feature-based logic ipss003 Correction of nonuniform illumination www.MatlabSite.com 11-9 a pplylut Purpose 11applylut Perform neighborhood operations on binary images, using lookup tables Synta x A = applylut(BW,lut) Description A = applylut(BW,lut) performs a 2-by-2 or 3-by-3 neighborhood operation on binary image BW by using a lookup table (lut). lut is either a 16-element or 512-element vector returned by makelut. The vector consists of the output values for all possible 2-by-2 or 3-by-3 neighborhoods. The values returned in A depend on the values in lut. For example, if lut consists of all 1’s and 0’s, A will be a binary image. Cla ss Support BW and lut can be of class uint8 or double. If the class of lut is uint8, or if the elements of lut are all integers between 0 and 255, then the class of A is uint8; otherwise, the class of A is double. Algorithm applylut performs a neighborhood operation on a binary image by producing a matrix of indices into lut, and then replacing the indices with the actual values in lut. The specific algorithm used depends on whether you use 2-by-2 or 3-by-3 neighborhoods. 2-by-2 neighborhoods For 2-by-2 neighborhoods, length(lut) is 16. There are 4 pixels in each neighborhood, and 2 possible states for each pixel, so the total number of permutations is 24 = 16. To produce the matrix of indices, applylut convolves the binary image BW with this matrix: 8 4 2 1 The resulting convolution contains integer values in the range [0,15]. applylut uses the central part of the convolution, of the same size as BW, and adds 1 to each value to shift the range to [1,16]. It then constructs A by replacing the values in the cells of the index matrix with the values in lut that the indices point to. 11 11-10 www.MatlabSite.com a pplylut 3-by-3 neighborhoods For 3-by-3 neighborhoods, length(lut) is 512. There are 9 pixels in each neighborhood, and 2 possible states for each pixel, so the total number of permutations is 29 = 512. To produce the matrix of indices, applylut convolves the binary image BW with this matrix: 256 128 64 32 16 8 4 2 1 The resulting convolution contains integer values in the range [0,511]. applylut uses the central part of the convolution, of the same size as BW, and adds 1 to each value to shift the range to [1,512]. It then constructs A by replacing the values in the cells of the index matrix with the values in lut that the indices point to. Ex a mple In this example, you perform erosion using a 2-by-2 neighborhood. An output pixel is on only if all four of the input pixel’s neighborhood pixels are on. lut = makelut('sum(x(:)) == 4',2); BW1 = imread('text.tif'); BW2 = applylut(BW1,lut); imshow(BW1) figure, imshow(BW2) See Also makelut www.MatlabSite.com 11-11 bestblk Purpose 11bestblk Determine block size for block processing Synta x siz = bestblk([m n],k) [mb,nb] = bestblk([m n],k) Description siz = bestblk([m n],k) returns, for an m-by-n image, the optimal block size for block processing. k is a scalar specifying the maximum row and column dimensions for the block; if the argument is omitted, it defaults to 100. siz is a 1-by-2 vector containing the row and column dimensions for the block. [mb,nb] = bestblk([m n],k) returns the row and column dimensions for the block in mb and nb, respectively. Algorithm bestblk returns the optimal block size given m, n, and k. The algorithm for determining siz is: • If m is less than or equal to k, return m. • If m is greater than k, consider all values between min(m/10,k/2) and k. Return the value that minimizes the padding required. The same algorithm is then repeated for n. Ex a mple siz = bestblk([640 800],72) siz = 64 See Also 11-12 50 blkproc www.MatlabSite.com blk proc Purpose 11blkproc Implement distinct block processing for an image Synta x B B B B Description B = blkproc(A,[m n],fun) processes the image A by applying the function fun to each distinct m-by-n block of A, padding A with zeros if necessary. fun can be an inline function, a text string containing the name of a function, or a string containing an expression. fun should operate on an m-by-n block, x, and return a matrix, vector, or scalar y: = = = = blkproc(A,[m n],fun) blkproc(A,[m n],fun,P1,P2,...) blkproc(A,[m n],[mborder nborder],fun,...) blkproc(A,'indexed',...) y = fun(x) blkproc does not require that y be the same size as x. However, B is the same size as A only if y is the same size as x. B = blkproc(A,[m n],fun,P1,P2,...) passes the additional parameters P1,P2,..., to fun. B = blkproc(A,[m n],[mborder nborder],fun,...) defines an overlapping border around the blocks. blkproc extends the original m-by-n blocks by mborder on the top and bottom, and nborder on the left and right, resulting in blocks of size (m+2*mborder)-by-(n+2*nborder). blkproc pads the border with zeros, if necessary, on the edges of A. fun should operate on the extended block. The line below processes an image matrix as 4-by-6 blocks, each having a row border of 2 and a column border of 3. Because each 4-by-6 block has this 2-by-3 border, fun actually operates on blocks of size 8-by-12. B = blkproc(A,[4 6],[2 3],fun,...) B = blkproc(A,'indexed',...) processes A as an indexed image, padding with zeros if the class of A is uint8, or ones if the class of A is double. Cla ss Support The input image A can be of any class supported by fun. The class of B depends on the class of the output from fun. www.MatlabSite.com 11-13 blk proc Ex a mple This example uses blkproc to set the pixels in each 8-by-8 block to the standard deviation of the elements in that block. I = imread('alumgrns.tif'); I2 = blkproc(I,[8 8],'std2(x)*ones(size(x))'); imshow(I) figure, imshow(I2,[]); See Also 11-14 colfilt, nlfilter www.MatlabSite.com brighten Purpose 11brighten Brighten or darken a colormap Synta x brighten(beta) newmap = brighten(beta) newmap = brighten(map,beta) brighten(fig,beta) Description brighten(beta) replaces the current colormap with a brighter or darker map that has essentially the same colors. The map is brighter if 0 < beta ≤ 1 and darker if –1 ≤ beta < 0. brighten(beta) followed by brighten(–beta) restores the original map. newmap = brighten(beta) returns a brighter or darker version of the current colormap without changing the display. newmap = brighten(map,beta) returns a brighter or darker version of the specified colormap without changing the display. brighten(fig,beta) brightens all of the objects in the figure fig. Rema rk s brighten is a function in MATLAB. See Also imadjust, rgbplot colormap in the online MATLAB Function Reference www.MatlabSite.com 11-15 bw a rea Purpose 11bwarea Compute the area of the objects in a binary image Synta x total = bwarea(BW) Description total = bwarea(BW) estimates the area of the objects in binary image BW. total is a scalar whose value corresponds roughly to the total number of on pixels in the image, but may not be exactly the same because different patterns of pixels are weighted differently. Cla ss Support BW can be of class uint8 or double. total is of class double. Algorithm bwarea estimates the area of all of the on pixels in an image by summing the areas of each pixel in the image. The area of an individual pixel is determined by looking at its 2-by-2 neighborhood. There are six different patterns distinguished, each representing a different area: • Patterns with zero on pixels (area = 0) • Patterns with one on pixel (area = 1/4) • Patterns with two adjacent on pixels (area = 1/2) • Patterns with two diagonal on pixels (area = 3/4) • Patterns with three on pixels (area = 7/8) • Patterns with all four on pixels (area = 1) Keep in mind that each pixel is part of four different 2-by-2 neighborhoods. This means, for example, that a single on pixel surrounded by off pixels has a total area of 1. 11-16 www.MatlabSite.com bw a rea Ex a mple This example computes the area in the objects of a 256-by-256 binary image. BW = imread('circles.tif'); imshow(BW); bwarea(BW) ans = 15799 See Also bweuler, bwperim References Pratt, William K. Digital Image Processing. New York: John Wiley & Sons, Inc., 1991. p. 634. www.MatlabSite.com 11-17 bw euler Purpose 11bweuler Compute the Euler number of a binary image Synta x eul = bweuler(BW,n) Description eul = bweuler(BW,n) returns the Euler number for the binary image BW. eul is a scalar whose value is the total number of objects in the image minus the total number of holes in those objects. n can have a value of either 4 or 8, where 4 specifies 4-connected objects and 8 specifies 8-connected objects; if the argument is omitted, it defaults to 8. Cla ss Support Ex a mple BW can be of class uint8 or double. eul is of class double. BW = imread('circles.tif'); imshow(BW); bweuler(BW) ans = –2 Algorithm bweuler computes the Euler number by considering patterns of convexity and concavity in local 2-by-2 neighborhoods. See Pratt, p. 633, for a discussion of the algorithm used. See Also 11-18 bwmorph, bwperim www.MatlabSite.com bw euler References Pratt, William K. Digital Image Processing. New York: John Wiley & Sons, Inc., 1991. p. 633. Horn, Berthold P. K., Robot Vision. New York: McGraw-Hill, 1986. pp. 73-77. www.MatlabSite.com 11-19 bw fill Purpose Synta x 11bwfill Fill background regions in a binary image BW2 = bwfill(BW1,c,r,n) BW2 = bwfill(BW1,n) [BW2,idx] = bwfill(...) BW2 = bwfill(x,y,BW1,xi,yi,n) [x,y,BW2,idx,xi,yi] = bwfill(...) BW2 = bwfill(BW1,'holes',n) [BW2,idx] = bwfill(BW1,'holes',n) Description BW2 = bwfill(BW1,c,r,n) performs a flood-fill operation on the input binary image BW1, starting from the pixel (r,c). If r and c are equal-length vectors, the fill is performed in parallel from the starting pixels (r(k),c(k)). n can have a value of either 4 or 8 (the default), where 4 specifies 4-connected foreground and 8 specifies 8-connected foreground. The foreground of BW1 comprises the on pixels (i.e., having value of 1). BW2 = bwfill(BW1,n) displays the image BW1 on the screen and lets you select the starting points using the mouse. If you omit BW1, bwfill operates on the image in the current axes. Use normal button clicks to add points. Press Backspace or Delete to remove the previously selected point. A shift-click, right-click, or double-click selects a final point and then starts the fill; pressing Return finishes the selection without adding a point. [BW2,idx] = bwfill(...) returns the linear indices of all pixels filled by bwfill. BW2 = bwfill(x,y,BW1,xi,yi,n) uses the vectors x and y to establish a nondefault spatial coordinate system for BW1. xi and yi are scalars or equal-length vectors that specify locations in this coordinate system. [x,y,BW2,idx,xi,yi] = bwfill(...) returns the XData and YData in x and y; the output image in BW2; linear indices of all filled pixels in idx; and the fill starting points in xi and yi. BW2 = bwfill(BW1,'holes',n) fills the holes in the binary image BW1. bwfill automatically determines which pixels are in object holes, and then changes the value of those pixels from 0 to 1. n defaults to 8 if you omit the argument. 11-20 www.MatlabSite.com bw fill [BW2,idx] = bwfill(BW1,'holes',n) returns the linear indices of all pixels filled in by bwfill. If bwfill is used with no output arguments, the resulting image is displayed in a new figure. Rema rk s bwfill differs from many other binary image operations in that it operates on background pixels, rather than foreground pixels. If the foreground is 8-connected, the background is 4-connected, and vice versa. Note, however, that you specify the connectedness of the foreground when you call bwfill. Cla ss Support The input image BW1 can be of class double or uint8. The output image BW2 is of class uint8. www.MatlabSite.com 11-21 bw fill Ex a mples BW1 = [1 1 1 1 1 1 1 1 0 1 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 1 0 0 1 1 0 0 0 1 1 1 0 1 1 1 0 0 0 1 1 0 0 1 0 0 1 1 1 1 1 1 0 0 0 0 1 0 0 0] BW2 = bwfill(BW1,3,3,8) BW2 = 1 1 1 1 1 1 1 1 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 1 0 0 0 1 1 1 0 1 1 1 I = imread('blood1.tif'); BW3 = ~im2bw(I); BW4 = bwfill(BW3,'holes'); imshow(BW3) figure, imshow(BW4) See Also 11-22 bwselect, roifill www.MatlabSite.com 0 0 0 1 1 0 0 1 0 0 1 1 1 1 1 1 0 0 0 0 1 0 0 0 bw la bel Purpose 11bwlabel Label connected components in a binary image Synta x L = bwlabel(BW,n) [L,num] = bwlabel(BW,n) Description L = bwlabel(BW,n) returns a matrix L, of the same size as BW, containing labels for the connected objects in BW. n can have a value of either 4 or 8, where 4 specifies 4-connected objects and 8 specifies 8-connected objects; if the argument is omitted, it defaults to 8. The elements of L are integer values greater than or equal to 0. The pixels labeled 0 are the background. The pixels labeled 1 make up one object, the pixels labeled 2 make up a second object, and so on. [L,num] = bwlabel(BW,n) returns in num the number of connected objects found in BW. Cla ss Support The input image BW can be of class double or uint8. The output matrix L is of class double. Rema rk s You can use the MATLAB find function in conjunction with bwlabel to return vectors of indices for the pixels that make up a specific object. For example, to return the coordinates for the pixels in object 2: [r,c] = find(bwlabel(BW)==2) You can display the output matrix as a pseudocolor indexed image. Each object appears in a different color, so the objects are easier to distinguish than in the original image. To do this, you must first add 1 to each element in the output matrix, so that the data is in the proper range. Also, it is good idea to use a colormap in which the first few colors are very distinct. www.MatlabSite.com 11-23 bw la bel Ex a mple This example illustrates using 4-connected objects. Notice objects 2 and 3; with 8-connected labeling, bwlabel would consider these a single object rather than two separate objects. BW = [1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 1 1 0 0 0 1 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0] L = bwlabel(BW,4) L = 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 2 2 0 0 0 0 0 0 2 2 0 0 0 3 0 0 0 0 3 3 3 3 0 0 0 0 0 0 0 0 0 [r,c] = find(L==2); rc = [r c] rc = 2 3 2 3 5 5 6 6 See Also bweuler, bwselect Reference Haralick, Robert M., and Linda G. Shapiro. Computer and Robot Vision, Volume I. Addison-Wesley, 1992. pp. 28-48. 11-24 www.MatlabSite.com bw m orph Purpose 11bwmorph Perform morphological operations on binary images Synta x BW2 = bwmorph(BW1,operation) BW2 = bwmorph(BW1,operation,n) Description BW2 = bwmorph(BW1,operation) applies a specific morphological operation to the binary image BW1. BW2 = bwmorph(BW1,operation,n) applies the operation n times. n can be Inf, in which case the operation is repeated until the image no longer changes. operation is a string that can have one of these values: 'bothat' 'erode' 'shrink' 'bridge' 'fill' 'skel' 'clean' 'hbreak' 'spur' 'close' 'majority' 'thicken' 'diag' 'open' 'thin' 'dilate' 'remove' 'tophat' 'bothat' (“bottom hat”) performs binary closure (dilation followed by erosion) and subtracts the original image. 'bridge' bridges previously unconnected pixels. For example: 1 1 0 0 0 0 0 1 1 becomes 1 1 0 0 1 0 0 1 1 'clean' removes isolated pixels (individual 1’s that are surrounded by 0’s), such as the center pixel in this pattern: 0 0 0 0 1 0 0 0 0 'close' performs binary closure (dilation followed by erosion). www.MatlabSite.com 11-25 bw morph 'diag' uses diagonal fill to eliminate 8-connectivity of the background. For example: 0 1 0 1 0 0 0 0 0 becomes 0 1 0 1 1 0 0 0 0 'dilate' performs dilation using the structuring element ones(3). 'erode' performs dilation using the structuring element ones(3). 'fill' fills isolated interior pixels (individual 0’s that are surrounded by 1’s), such as the center pixel in this pattern: 1 1 1 1 0 1 1 1 1 'hbreak' removes H-connected pixels. For example: 1 0 1 1 1 1 1 0 1 becomes 1 0 1 1 0 1 1 0 1 'majority' sets a pixel to 1 if five or more pixels in its 3-by-3 neighborhood are 1’s; otherwise, it sets the pixel to 0. 'open' implements binary opening (erosion followed by dilation). 'remove' removes interior pixels. This option sets a pixel to 0 if all of its 4-connected neighbors are 1, thus leaving only the boundary pixels on. 'shrink', with n = Inf, shrinks objects to points. It removes pixels so that objects without holes shrink to a point, and objects with holes shrink to a connected ring halfway between each hole and the outer boundary. This option preserves the Euler number. 'skel', with n = Inf, removes pixels on the boundaries of objects but does not allow objects to break apart. The pixels remaining make up the image skeleton. This option preserves the Euler number. 11-26 www.MatlabSite.com bw m orph 'spur' removes spur pixels. For example: 0 0 0 0 1 0 0 0 1 1 0 0 1 0 0 0 0 0 0 0 becomes 0 0 0 0 1 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 'thicken', with n = Inf, thickens objects by adding pixels to the exterior of objects until doing so would result in previously unconnected objects being 8-connected. This option preserves the Euler number. 'thin', with n = Inf, thins objects to lines. It removes pixels so that an object without holes shrinks to a minimally connected stroke, and an object with holes shrinks to a connected ring halfway between each hole and the outer boundary. This option preserves the Euler number. 'tophat' (“top hat”) returns the image minus the binary opening of the image. Cla ss Support The input image BW1 can be of class double or uint8. The output image BW2 is of class uint8. www.MatlabSite.com 11-27 bw morph Ex a mples BW1 = imread('circles.tif'); imshow(BW1); BW2 = bwmorph(BW1,'remove'); BW3 = bwmorph(BW1,'skel',Inf); imshow(BW2) figure, imshow(BW3) See Also bweuler, bwperim, dilate, erode References Haralick, Robert M., and Linda G. Shapiro. Computer and Robot Vision, Volume I. Addison-Wesley, 1992. Pratt, William K. Digital Image Processing. John Wiley & Sons, Inc., 1991. 11-28 www.MatlabSite.com bw perim Purpose 11bwperim Determine the perimeter of the objects in a binary image Synta x BW2 = bwperim(BW1,n) Description BW2 = bwperim(BW1,n) returns a binary image containing only the perimeter pixels of objects in the input image BW1. A pixel is part of the perimeter if its value is 1 and there is at least one zero-valued pixel in its neighborhood. n can have a value of either 4 or 8, where 4 specifies 4-connected neighborhoods and 8 specifies 8-connected neighborhoods; if the argument is omitted, it defaults to 8. Cla ss Support The input image BW1 can be of class double or uint8. The output image BW2 is of class uint8. Ex a mple See Also BW1 = imread('circbw.tif'); BW2 = bwperim(BW1,8); imshow(BW1) figure, imshow(BW2) bwarea, bweuler, bwfill www.MatlabSite.com 11-29 bw select Purpose Synta x 11bwselect Select objects in a binary image BW2 = bwselect(BW1,c,r,n) BW2 = bwselect(BW1,n) [BW2,idx] = bwselect(...) BW2 = bwselect(x,y,BW1,xi,yi,n) [x,y,BW2,idx,xi,yi] = bwselect(...) Description BW2 = bwselect(BW1,c,r,n) returns a binary image containing the objects that overlap the pixel (r,c). r and c can be scalars or equal-length vectors. If r and c are vectors, BW2 contains the sets of objects overlapping with any of the pixels (r(k),c(k)). n can have a value of either 4 or 8 (the default), where 4 specifies 4-connected objects and 8 specifies 8-connected objects. Objects are connected sets of on pixels (i.e., pixels having a value of 1). BW2 = bwselect(BW1,n) displays the image BW1 on the screen and lets you select the (r,c) coordinates using the mouse. If you omit BW1, bwselect operates on the image in the current axes. Use normal button clicks to add points. Pressing Backspace or Delete removes the previously selected point. A shift-click, right-click, or double-click selects the final point; pressing Return finishes the selection without adding a point. [BW2,idx] = bwselect(...) returns the linear indices of the pixels belonging to the selected objects. BW2 = bwselect(x,y,BW1,xi,yi,n) uses the vectors x and y to establish a nondefault spatial coordinate system for BW1. xi and yi are scalars or equal-length vectors that specify locations in this coordinate system. [x,y,BW2,idx,xi,yi] = bwselect(...) returns the XData and YData in x and y; the output image in BW2; linear indices of all the pixels belonging to the selected objects in idx; and the specified spatial coordinates in xi and yi. If bwselect is called with no output arguments, the resulting image is displayed in a new figure. 11-30 www.MatlabSite.com bw select Ex a mple BW1 = imread('text.tif'); c = [16 90 144]; r = [85 197 247]; BW2 = bwselect(BW1,c,r,4); imshow(BW1) figure, imshow(BW2) Cla ss Support The input image BW1 can be of class double or uint8. The output image BW2 is of class uint8. See Also bwfill, bwlabel, impixel, roipoly, roifill www.MatlabSite.com 11-31 cmpermute Purpose 11cmpermute Rearrange the colors in a colormap Synta x [Y,newmap] = cmpermute(X,map) [Y,newmap] = cmpermute(X,map,index) Description [Y,newmap] = cmpermute(X,map) randomly reorders the colors in map to produce a new colormap newmap. cmpermute also modifies the values in X to maintain correspondence between the indices and the colormap, and returns the result in Y. The image Y and associated colormap newmap produce the same image as X and map. [Y,newmap] = cmpermute(X,map,index) uses an ordering matrix (such as the second output of sort) to define the order of colors in the new colormap. Cla ss Support The input image X can be of class uint8 or double. Y is returned as an array of the same class as X. Ex a mple To order a colormap by luminance, use: ntsc = rgb2ntsc(map); [dum,index] = sort(ntsc(:,1)); [Y,newmap] = cmpermute(X,map,index); See Also 11-32 randperm, sort in the online MATLAB Function Reference www.MatlabSite.com cm unique Purpose 11cmunique Find unique colormap colors and the corresponding image Synta x [Y,newmap] = cmunique(X,map) [Y,newmap] = cmunique(RGB) [Y,newmap] = cmunique(I) Description [Y,newmap] = cmunique(X,map) returns the indexed image Y and associated colormap newmap that produce the same image as (X,map) but with the smallest possible colormap. cmunique removes duplicate rows from the colormap and adjusts the indices in the image matrix accordingly. [Y,newmap] = cmunique(RGB) converts the truecolor image RGB to the indexed image Y and its associated colormap newmap. newmap is the smallest possible colormap for the image, containing one entry for each unique color in RGB. (Note that newmap may be very large, because the number of entries can be as many as the number of pixels in RGB.) [Y,newmap] = cmunique(I) converts the intensity image I to an indexed image Y and its associated colormap newmap. newmap is the smallest possible colormap for the image, containing one entry for each unique intensity level in I. Cla ss Support The input image can be of class uint8 or double. The class of the output image Y is uint8 if the length of newmap is less than or equal to 256. If the length of newmap is greater than 256, Y is of class double. See Also gray2ind, rgb2ind www.MatlabSite.com 11-33 col2 im Purpose 11col2im Rearrange matrix columns into blocks Synta x A = col2im(B,[m n],[mm nn],block_type) A = col2im(B,[m n],[mm nn]) Description col2im rearranges matrix columns into blocks. block_type is a string with one of these values: • 'distinct' for m-by-n distinct blocks • 'sliding' for m-by-n sliding blocks (default) A = col2im(B,[m n],[mm nn],'distinct') rearranges each column of B into a distinct m-by-n block to create the matrix A of size mm-by-nn. If B = [A11(:) A12(:) A21(:) A22(:)], where each column has length m*n, then A = [A11 A12;A21 A22] where each Aij is m-by-n. A = col2im(B,[m n],[mm nn],'sliding') rearranges the row vector B into a matrix of size (mm–m+1)-by-(nn–n+1). B must be a vector of size 1-by-(mm–m+1)*(nn–n+1). B is usually the result of processing the output of im2col(...,'sliding') using a column compression function (such as sum). A = col2im(B,[m n],[mm nn]) uses the default block_type of 'sliding'. Cla ss Support B can be of class uint8 or double. A is of the same class as B. See Also blkproc, colfilt, im2col, nlfilter 11-34 www.MatlabSite.com colfilt Purpose 11colfilt Perform neighborhood operations using columnwise functions Synta x B B B B Description colfilt processes distinct or sliding blocks as columns. colfilt can perform similar operations to blkproc and nlfilter, but often executes much faster. = = = = colfilt(A,[m n],block_type,fun) colfilt(A,[m n],block_type,fun,P1,P2,...) colfilt(A,[m n],[mblock nblock],block_type,fun,...) colfilt(A,'indexed',...) B = colfilt(A,[m n],block_type,fun) processes the image A by rearranging each m-by-n block of A into a column of a temporary matrix, and then applying the function fun to this matrix. fun can be an inline function, a text string containing the name of a function, or a string containing an expression. colfilt zero pads A, if necessary. Before calling fun, colfilt calls im2col to create the temporary matrix. After calling fun, colfilt rearranges the columns of the matrix back into m-by-n blocks using col2im. block_type is a string with one of these values: • 'distinct' for m-by-n distinct blocks • 'sliding' for m-by-n sliding neighborhoods B = colfilt(A,[m n],'distinct',fun) rearranges each m-by-n distinct block of A into a column in a temporary matrix, and then applies the function fun to this matrix. fun must return a matrix of the same size as the temporary matrix. colfilt then rearranges the columns of the matrix returned by fun into m-by-n distinct blocks. B = colfilt(A,[m n],'sliding',fun) rearranges each m-by-n sliding neighborhood of A into a column in a temporary matrix, and then applies the function fun to this matrix. fun must return a row vector containing a single value for each column in the temporary matrix. (Column compression functions such as sum return the appropriate type of output.) colfilt then rearranges the vector returned by fun into a matrix of the same size as A. www.MatlabSite.com 11-35 colfilt B = colfilt(A,[m n],block_type,fun,P1,P2,...) passes the additional parameters P1,P2,..., to fun. colfilt calls fun using: y = fun(x,P1,P2,...) where x is the temporary matrix before processing, and y is the temporary matrix after processing. B = colfilt(A,[m n],[mblock nblock],block_type,fun,...) processes the matrix A as above, but in blocks of size mblock-by-nblock to save memory. Note that using the [mblock nblock] argument does not change the result of the operation. B = colfilt(A,'indexed',...) processes A as an indexed image, padding with zeros if the class of A is uint8, or ones if the class of A is double. Cla ss Support The input image A can be of any class supported by fun. The class of B depends on the class of the output from fun. Ex a mple This example sets each output pixel to the mean value of the input pixel’s 5-by-5 neighborhood: I2 = colfilt(I,[5 5],'sliding','mean'); See Also 11-36 blkproc, col2im, im2col, nlfilter www.MatlabSite.com colorba r Purpose 11colorbar Display a colorbar Synta x colorbar('vert') colorbar('horiz') colorbar(h) colorbar h = colorbar(...) Description colorbar('vert') appends a vertical colorbar to the current axes, resizing the axes to make room for the colorbar. colorbar works with both two-dimensional and three-dimensional plots. colorbar('horiz') appends a horizontal colorbar to the current axes. colorbar(h) places the colorbar in the axes h. The colorbar is horizontal if the width of the axes is greater than its height. colorbar with no arguments adds a new vertical colorbar or updates an existing one. h = colorbar(...) returns a handle to the colorbar axes. Rema rk s colorbar is a function in MATLAB. www.MatlabSite.com 11-37 colorba r Ex a mples Display a colorbar to view values for a filtered image. I = imread('blood1.tif'); h = fspecial('log'); I2 = filter2(h,I); imshow(I2,[]), colormap(jet(64)), colorbar 11-38 www.MatlabSite.com conv2 Purpose 11conv2 Perform two-dimensional convolution Synta x C = conv2(A,B) C = conv2(hcol,hrow,A) C = conv2(...,shape) Description C = conv2(A,B) performs the two-dimensional convolution of matrices A and B, returning the result in the output matrix C. The size in each dimension of C is equal to the sum of the corresponding dimensions of the input matrices minus one. That is, if the size of A is [ma,mb] and the size of B is [mb,nb], then the size of C is [ma+mb–1,na+nb–1]. C = conv2(hcol,hrow,A) convolves A separably with hcol in the column direction and hrow in the row direction. hcol and hrow are both vectors. C = conv2(...,shape) returns a subsection of the two-dimensional convolution, as specified by the shape parameter. shape is a string with one of these values: • 'full' (the default) returns the full two-dimensional convolution. • 'same' returns the central part of the convolution of the same size as A. • 'valid' returns only those parts of the convolution that are computed without the zero-padded edges. Using this option, size(C) = [ma–mb+1, na–nb+1] when size(A) > size(B). For image filtering, A should be the image matrix and B should be the filter (convolution kernel) if the shape parameter is 'same' or 'valid'. If the shape parameter is 'full', the order does not matter, because full convolution is commutative. Cla ss Support All vector and matrix inputs to conv2 can be of class uint8 or double. The output matrix C is of class double. Rema rk s conv2 is a function in MATLAB. www.MatlabSite.com 11-39 conv2 Ex a mple A = magic(5) A = 17 23 4 10 11 24 5 6 12 18 1 7 13 19 25 8 14 20 21 2 15 16 22 3 9 B = [1 2 1;0 2 0;3 1 3] B = 1 0 3 2 2 1 1 0 3 C = conv2(A,B) C = 17 23 55 79 23 30 33 See Also 58 85 149 78 82 68 65 66 88 117 160 153 135 126 34 35 163 161 199 168 85 32 67 159 187 205 91 104 38 76 135 129 108 84 15 15 16 67 51 75 9 27 filter2 xcorr, xcorr2 in the Signal Processing Toolbox User’s Guide conv, deconv in the online MATLAB Function Reference 11-40 www.MatlabSite.com convmtx 2 Purpose 11convmtx2 Compute two-dimensional convolution matrix Synta x T = convmtx2(H,m,n) T = convmtx2(H,[m n]) Description T = convmtx2(H,m,n) or T = convmtx2(H,[m n]) returns the convolution matrix T for the matrix H. If X is an m-by-n matrix, then reshape(T*X(:),size(H)+[m n]–1) is the same as conv2(X,H). T is returned as a sparse matrix. The number of nonzero elements in T is no larger than prod(size(H))*m*n. See Also conv2 convmtx in the Signal Processing Toolbox User’s Guide www.MatlabSite.com 11-41 convn Purpose 11convn Perform N-dimensional convolution Synta x C = convn(A,B) C = convn(A,B,shape) Description C = convn(A,B) computes the N-dimensional convolution of matrices A and B. C = convn(A,B,shape) returns a subsection of the N-dimensional convolution, as specified by the shape parameter. shape is a string with one of these values: • 'full' (the default) returns the full convolution. • 'same' returns the central part of the convolution of the same size as A. • 'valid' returns only those parts of the convolution that are computed without zero-padded edges. Cla ss Support The input matrices A and B can be of class uint8 or double. The output matrix C is of class double. Rema rk s convn is a function in MATLAB. See Also conv2 11-42 www.MatlabSite.com corr2 Purpose 11corr2 Compute the two-dimensional correlation coefficient between two matrices Synta x r = corr2(A,B) Description r = corr2(A,B) computes the correlation coefficient between A and B, where A and B are matrices or vectors of the same size. Cla ss Support A and B can be of class uint8 or double. r is a scalar of class double. Algorithm corr2 computes the correlation coefficient using ∑ ∑ ( Amn – A ) ( Bmn – B ) m n r = -----------------------------------------------------------------------------------------------------2  2  ( B mn – B ) ( A mn – A )    ∑∑ m n ∑∑ m n where A = mean2(A), and B = mean2(B). See Also std2 corrcoef in the online MATLAB Function Reference www.MatlabSite.com 11-43 dct2 Purpose 11dct2 Compute two-dimensional discrete cosine transform Synta x B = dct2(A) B = dct2(A,m,n) B = dct2(A,[m n]) Description B = dct2(A) returns the two-dimensional discrete cosine transform of A. The matrix B is the same size as A and contains the discrete cosine transform coefficients B(k1,k2). B = dct2(A,m,n) or B = dct2(A,[m n]) pads the matrix A with zeros to size m-by-n before transforming. If m or n is smaller than the corresponding dimension of A, dct2 truncates A. Cla ss Support A can be of class uint8 or double. The returned matrix B is of class double. Algorithm The discrete cosine transform (DCT) is closely related to the discrete Fourier transform. It is a separable, linear transformation; that is, the two-dimensional transform is equivalent to a one-dimensional DCT performed along a single dimension followed by a one-dimensional DCT in the other dimension. The definition of the two-dimensional DCT for an input image A and output image B is M–1 N–1 B pq = α p α q ∑ ∑ m=0 n=0 π ( 2m + 1 ) p π ( 2n + 1 )q A mn cos ------------------------------- cos ----------------------------- , 2M 2N  1 ⁄ M, p = 0 αp =   2 ⁄ M, 1 ≤ p ≤ M – 1 0≤ p≤M–1 0≤q≤N–1 1 ⁄ N , q = 0 αq =   2 ⁄ N, 1 ≤ q ≤ N – 1 where M and N are the row and column size of A, respectively. If you apply the DCT to real data, the result is also real. The DCT tends to concentrate information, making it useful for image compression applications. This transform can be inverted using idct2. 11-44 www.MatlabSite.com dct2 Ex a mple The commands below compute the discrete cosine transform for the autumn image. Notice that most of the energy is in the upper-left corner. RGB = imread('autumn.tif'); I = rgb2gray(RGB); J = dct2(I); imshow(log(abs(J)),[]), colormap(jet(64)), colorbar Now set values less than magnitude 10 in the DCT matrix to zero, and then reconstruct the image using the inverse DCT function idct2. J(abs(J) < 10) = 0; K = idct2(J)/255; imshow(K) See Also fft2, idct2, ifft2 References Jain, Anil K. Fundamentals of Digital Image Processing. Englewood Cliffs, NJ: Prentice Hall, 1989. pp. 150-153. Pennebaker, William B., and Joan L. Mitchell. JPEG: Still Image Data Compression Standard. Van Nostrand Reinhold, 1993. www.MatlabSite.com 11-45 dctmtx Purpose 11dctmtx Compute discrete cosine transform matrix Synta x D = dctmtx(n) Description D = dctmtx(n) returns the n-by-n DCT transform matrix. D*A is the DCT of the columns of A and D'*A is the inverse DCT of the columns of A (when A is n-by-n). Cla ss Support n is a scalar of class double. D is returned as a matrix of class double. Rema rk s If A is square, the two-dimensional DCT of A can be computed as D*A*D'. This computation is sometimes faster than using dct2, especially if you are computing a large number of small DCTs, because D needs to be determined only once. For example, in JPEG compression, the DCT of each 8-by-8 block is computed. To perform this computation, use dctmtx to determine D, and then calculate each DCT using D*A*D' (where A is each 8-by-8 block). This is faster than calling dct2 for each individual block. See Also 11-46 dct2 www.MatlabSite.com dila te Purpose 11dilate Perform dilation on a binary image Synta x BW2 = dilate(BW1,SE) BW2 = dilate(BW1,SE,alg) BW2 = dilate(BW1,SE,...,n) Description BW2 = dilate(BW1,SE) performs dilation on the binary image BW1, using the binary structuring element SE. SE is a matrix containing only 1’s and 0’s. BW2 = dilate(BW1,SE,alg) performs dilation using the specified algorithm. alg is a string that can have one of these values: • 'spatial' (default) – processes the image in the spatial domain. • 'frequency' – processes the image in the frequency domain. Both algorithms produce the same result, but they make different tradeoffs between speed and memory use. The frequency algorithm is faster for large images and structuring elements than the spatial algorithm, but uses much more memory. BW2 = dilate(BW1,SE,...,n) performs the dilation operation n times. Cla ss Support The input image BW1 can be of class double or uint8. The output image BW2 is of class uint8. Rema rk s You should use the frequency algorithm only if you have a large amount of memory on your system. If you use this algorithm with insufficient memory, it may actually be slower than the spatial algorithm, due to virtual memory paging. If the frequency algorithm slows down your system excessively, or if you receive “out of memory” messages, use the spatial algorithm instead. www.MatlabSite.com 11-47 dila te Ex a mple BW1 = imread('text.tif'); SE = ones(6,2); BW2 = dilate(BW1,SE); imshow(BW1) figure, imshow(BW2) See Also bwmorph, erode References Haralick, Robert M., and Linda G. Shapiro. Computer and Robot Vision, Volume I. Addison-Wesley, 1992. p. 158. Gonzalez, Rafael C., and Richard E. Woods. Digital Image Processing. Addison-Wesley, 1992. p. 518. 11-48 www.MatlabSite.com dither Purpose 11dither Convert an image, increasing apparent color resolution by dithering Synta x X = dither(RGB,map) BW = dither(I) Description X = dither(RGB,map) creates an indexed image approximation of the RGB image in the array RGB by dithering the colors in colormap map. X = dither(RGB,map,Qm,Qe) creates an indexed image from RGB, specifying the parameters Qm and Qe. Qm specifies the number of quantization bits to use along each color axis for the inverse color map, and Qe specifies the number of quantization bits to use for the color space error calculations. If Qe < Qm, dithering cannot be performed and an undithered indexed image is returned in X. If you omit these parameters, dither uses the default values Qm = 5, Qe = 8. BW = dither(I) converts the intensity image in the matrix I to the binary (black and white) image BW by dithering. Cla ss Support The input image (RGB or I) can be of class double or uint8. All other input arguments must be of class double. The output image (X or BW) is of class uint8 if it is a binary image or if it is an indexed image with 256 or fewer colors; otherwise its class is double. Algorithm dither increases the apparent color resolution of an image by applying Floyd-Steinberg’s error diffusion dither algorithm. Reference Floyd, R. W. and L. Steinberg. “An Adaptive Algorithm for Spatial Gray Scale.” International Symposium Digest of Technical Papers. Society for Information Displays, 1975. p. 36. Lim, Jae S. Two-Dimensional Signal and Image Processing. Englewood Cliffs, NJ: Prentice Hall, 1990. pp. 469-476. See Also rgb2ind www.MatlabSite.com 11-49 double Purpose 11double Convert data to double precision Synta x B = double(A) Description B = double(A) creates a double-precision array B from the array A. If A is a double array, B is identical to A. double is useful if you have a uint8 image array that you want to perform arithmetic operations on, because MATLAB does not support these operations on uint8 data. Rema rk s Ex a mple See Also 11-50 double is a MATLAB built-in function. A = imread('saturn.tif'); B = sqrt(double(A)); im2double, im2uint, uint8 www.MatlabSite.com edge Purpose Synta x 11edge Find edges in an intensity image BW = edge(I,'sobel') BW = edge(I,'sobel',thresh) BW = edge(I,'sobel',thresh,direction) [BW,thresh] = edge(I,'sobel',...) BW = edge(I,'prewitt') BW = edge(I,'prewitt',thresh) BW = edge(I,'prewitt',thresh,direction) [BW,thresh] = edge(I,'prewitt',...) BW = edge(I,'roberts') BW = edge(I,'roberts',thresh) [BW,thresh] = edge(I,'roberts',...) BW = edge(I,'log') BW = edge(I,'log',thresh) BW = edge(I,'log',thresh,sigma) [BW,threshold] = edge(I,'log',...) BW = edge(I,'zerocross',thresh,h) [BW,thresh] = edge(I,'zerocross',...) BW = edge(I,'canny') BW = edge(I,'canny',thresh) BW = edge(I,'canny',thresh,sigma) [BW,threshold] = edge(I,'canny',...) Description edge takes an intensity image I as its input, and returns a binary image BW of the same size as I, with 1’s where the function finds edges in I and 0’s elsewhere. 11 www.MatlabSite.com 11-51 edge edge supports six different edge-finding methods: • The Sobel method finds edges using the Sobel approximation to the derivative. It returns edges at those points where the gradient of I is maximum. • The Prewitt method finds edges using the Prewitt approximation to the derivative. It returns edges at those points where the gradient of I is maximum. • The Roberts method finds edges using the Roberts approximation to the derivative. It returns edges at those points where the gradient of I is maximum. • The Laplacian of Gaussian method finds edges by looking for zero crossings after filtering I with a Laplacian of Gaussian filter. • The zero-cross method finds edges by looking for zero crossings after filtering I with a filter you specify. • The Canny method finds edges by looking for local maxima of the gradient of I. The gradient is calculated using the derivative of a Gaussian filter. The method uses two thresholds, to detect strong and weak edges, and includes the weak edges in the output only if they are connected to strong edges. This method is therefore less likely than the others to be “fooled” by noise, and more likely to detect true weak edges. The parameters you can supply differ depending on the method you specify. If you do not specify a method, edge uses the Sobel method. Sobel Method BW = edge(I,'sobel') specifies the Sobel method. BW = edge(I,'sobel',thresh) specifies the sensitivity threshold for the Sobel method. edge ignores all edges that are not stronger than thresh. If you do not specify thresh, or if thresh is empty ([]), edge chooses the value automatically. BW = edge(I,'sobel',thresh,direction) specifies directionality for the Sobel method. direction is a string specifying whether to look for 'horizontal' or 'vertical' edges, or 'both' (the default). [BW,thresh] = edge(I,'sobel',...) returns the threshold value. 11-52 www.MatlabSite.com edge Prewitt Method BW = edge(I,'prewitt') specifies the Prewitt method. BW = edge(I,'prewitt',thresh) specifies the sensitivity threshold for the Prewitt method. edge ignores all edges that are not stronger than thresh. If you do not specify thresh, or if thresh is empty ([]), edge chooses the value automatically. BW = edge(I,'prewitt',thresh,direction) specifies directionality for the Prewitt method. direction is a string specifying whether to look for 'horizontal' or 'vertical' edges, or 'both' (the default). [BW,thresh] = edge(I,'prewitt',...) returns the threshold value. Roberts Method BW = edge(I,method) specifies the Roberts method. BW = edge(I,method,thresh) specifies the sensitivity threshold for the Roberts method. edge ignores all edges that are not stronger than thresh. If you do not specify thresh, or if thresh is empty ([]), edge chooses the value automatically. [BW,thresh] = edge(I,method,...) returns the threshold value. Laplacian of Gaussian Method BW = edge(I,'log') specifies the Laplacian of Gaussian method. BW = edge(I,'log',thresh) specifies the sensitivity threshold for the Laplacian of Gaussian method. edge ignores all edges that are not stronger than thresh. If you do not specify thresh, or if thresh is empty ([]), edge chooses the value automatically. BW = edge(I,'log',thresh,sigma) specifies the Laplacian of Gaussian method, using sigma as the standard deviation of the LoG filter. The default sigma is 2; the size of the filter is n-by-n, where n = ceil(sigma*3)*2+1. [BW,thresh] = edge(I,'log',...) returns the threshold value. www.MatlabSite.com 11-53 edge Zero-cross Method BW = edge(I,'zerocross',thresh,h) specifies the zero-cross method, using the filter h. thresh is the sensitivity threshold; if the argument is empty ([]), edge chooses the sensitivity threshold automatically. [BW,thresh] = edge(I,'zerocross',...) returns the threshold value. Canny Method BW = edge(I,'canny') specifies the Canny method. BW = edge(I,'canny',thresh) specifies sensitivity thresholds for the Canny method. thresh is a two-element vector in which the first element is the low threshold, and the second element is the high threshold. If you specify a scalar for thresh, this value is used for the high threshold and 0.4*thresh is used for the low threshold. If you do not specify thresh, or if thresh is empty ([]), edge chooses low and high values automatically. BW = edge(I,'canny',thresh,sigma) specifies the Canny method, using sigma as the standard deviation of the Gaussian filter. The default sigma is 1; the size of the filter is chosen automatically, based on sigma. [BW,thresh] = edge(I,'canny',...) returns the threshold values as a two-element vector. Cla ss Support I can be of class uint8 or double. BW is of class uint8. Rema rk s For the 'log' and 'zerocross' methods, if you specify a threshold of 0, the output image has closed contours, because it includes all of the zero crossings in the input image. 11-54 www.MatlabSite.com edge Ex a mple Find the edges of the rice.tif image using the Prewitt and Canny methods: I = imread('rice.tif'); BW1 = edge(I,'prewitt'); BW2 = edge(I,'canny'); imshow(BW1) figure, imshow(BW2) References Lim, Jae S. Two-Dimensional Signal and Image Processing. Englewood Cliffs, NJ: Prentice Hall, 1990. pp. 478-488. Parker, James R. Algorithms for Image Processing and Computer Vision. New York: John Wiley & Sons, Inc., 1997. pp. 23-29. Canny, John. “A Computational Approach to Edge Detection,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 1986. Vol. PAMI-8, No. 6, pp. 679-698. www.MatlabSite.com 11-55 erode Purpose 11erode Perform erosion on a binary image Synta x BW2 = erode(BW1,SE) BW2 = erode(BW1,SE,alg) BW2 = erode(BW1,SE,...,n) Description BW2 = erode(BW1,SE) performs erosion on the binary image BW1, using the binary structuring element SE. SE is a matrix containing only 1’s and 0’s. BW2 = erode(BW1,SE,alg) performs erosion using the specified algorithm. alg is a string that can have one of these values: • 'spatial' (default) – processes the image in the spatial domain. • 'frequency' – processes the image in the frequency domain. Both algorithms produce the same result, but they make different tradeoffs between speed and memory use. The frequency algorithm is faster for large images and structuring elements than the spatial algorithm, but uses much more memory. BW2 = erode(BW1,SE,...,n) performs the erosion operation n times. Cla ss Support The input image BW1 can be of class double or uint8. The output image BW2 is of class uint8. Rema rk s You should use the frequency algorithm only if you have a large amount of memory on your system. If you use this algorithm with insufficient memory, it may actually be slower than the spatial algorithm, due to virtual memory paging. If the frequency algorithm slows down your system excessively, or if you receive “out of memory” messages, use the spatial algorithm instead. 11-56 www.MatlabSite.com erode Ex a mple BW1 = imread('text.tif') SE = ones(3,1) SE = 1 1 1 BW2 = erode(BW1,SE); imshow(BW1) figure, imshow(BW2) See Also bwmorph, dilate References Haralick, Robert M., and Linda G. Shapiro. Computer and Robot Vision, Volume I. Addison-Wesley, 1992. p. 158. Gonzalez, Rafael C., and Richard E. Woods. Digital Image Processing. Addison-Wesley, 1992. p. 518. www.MatlabSite.com 11-57 fft2 Purpose 11fft2 Compute two-dimensional fast Fourier transform (FFT) Synta x B = fft2(A) B = fft2(A,m,n) Description B = fft2(A) performs a two-dimensional fast Fourier transform, returning the result in B. B is the same size as A; if A is a vector, B has the same orientation as A. B = fft2(A,m,n) truncates or zero pads A, if necessary, to create an m-by-n matrix before performing the FFT. The result B is also m-by-n. Cla ss Support The input matrix A can be of class uint8 or double. The output matrix B is of class double. Rema rk s fft2 is a function in MATLAB. 11-58 www.MatlabSite.com fft2 Ex a mple load imdemos saturn2 imshow(saturn2) B = fftshift(fft2(saturn2)); imshow(log(abs(B)),[]), colormap(jet(64)), colorbar Algorithm fft2(A) is simply fft(fft(A).').' This computes the one-dimensional fft of each column A, then of each row of the result. The time required to compute fft2(A) depends on the number of prime factors of m and n. fft2 is fastest when m and n are powers of 2. See Also dct2, fftshift, idct2, ifft2 fft, ifft in the online MATLAB Function Reference www.MatlabSite.com 11-59 fftn Purpose 11fftn Compute N-dimensional fast Fourier transform Synta x B = fftn(A) B = fftn(A,siz) Description B = fftn(A) performs the N-dimensional fast Fourier transform. The result B is the same size as A. B = fftn(A,siz) pads A with zeros (or truncates A) to create an N-dimensional array of size siz before doing the transform. The size of the result is siz. Cla ss Support The input matrix A can be of class uint8 or double. The output matrix B is of class double. Rema rk s fftn is a function in MATLAB. Algorithm fftn(A) is equivalent to: B = A; for p = 1:length(size(A)) B = fft(B,[],p); end This code computes the one-dimensional fast Fourier transform along each dimension of A. The time required to compute fftn(A) depends strongly on the number of prime factors of the dimensions of A. It is fastest when all of the dimensions are powers of 2. See Also fft2, ifftn fft in the online MATLAB Function Reference 11-60 www.MatlabSite.com fftshift Purpose 11fftshift Shift DC component of fast Fourier transform to center of spectrum Synta x B = fftshift(A) Description B = fftshift(A) rearranges the outputs of fft, fft2, and fftn by moving the zero frequency component to the center of the array. For vectors, fftshift(A) swaps the left and right halves of A. For matrices, fftshift(A) swaps quadrants one and three of A with quadrants two and four. For higher-dimensional arrays, fftshift(A) swaps “half-spaces” of A along each dimension. Rema rk s Ex a mple See Also fftshift is a function in MATLAB. B = fftn(A); C = fftshift(B); fft2, fftn fft in the online MATLAB Function Reference www.MatlabSite.com 11-61 filter2 Purpose 11filter2 Perform two-dimensional linear filtering Synta x B = filter2(h,A) B = filter2(h,A,shape) Description B = filter2(h,A) filters the data in A with the two-dimensional FIR filter in the matrix h. It computes the result, B, using two-dimensional correlation, and returns the central part of the correlation that is the same size as A. B = filter2(h,A,shape) returns the part of B specified by the shape parameter. shape is a string with one of these values: • 'full' returns the full two-dimensional correlation. In this case, B is larger than A. • 'same' (the default) returns the central part of the correlation. In this case, B is the same size as A. • 'valid' returns only those parts of the correlation that are computed without zero-padded edges. In this case, B is smaller than A. Cla ss Support The matrix inputs to filter2 can be of class uint8 or double. The output matrix B is of class double. Rema rk s Two-dimensional correlation is equivalent to two-dimensional convolution with the filter matrix rotated 180 degrees. See the Algorithm section for more information about how filter2 performs linear filtering. filter2 is a function in MATLAB. 11-62 www.MatlabSite.com filter2 Ex a mple A = magic(6) A = 35 3 31 8 30 4 1 32 9 28 5 36 6 7 2 33 34 29 26 21 22 17 12 13 19 23 27 10 14 18 24 25 20 15 16 11 h = fspecial('sobel') h = 1 0 –1 2 0 –2 1 0 –1 B = filter2(h,A,'valid') B = –8 –23 –23 –8 Algorithm 4 –44 –50 4 4 –5 1 4 –8 40 40 –8 Given an image A and a two-dimensional FIR filter h, filter2 rotates your filter matrix (the computational molecule) 180 degrees to create a convolution kernel. It then calls conv2, the two-dimensional convolution function, to implement the filtering operation. filter2 uses conv2 to compute the full two-dimensional convolution of the FIR filter with the input matrix. By default, filter2 then extracts the central part of the convolution that is the same size as the input matrix, and returns this as the result. If the shape parameter specifies an alternate part of the convolution for the result, filter2 returns the appropriate part. See Also conv2, roifilt2 www.MatlabSite.com 11-63 freqspa ce Purpose 11freqspace Determine frequency spacing for two-dimensional frequency response Synta x [f1,f2] = freqspace(n) [f1,f2] = freqspace([m n]) [x1,y1] = freqspace(...,'meshgrid') f = freqspace(N) f = freqspace(N,'whole') Description freqspace returns the implied frequency range for equally spaced frequency responses. freqspace is useful when creating desired frequency responses for fsamp2, fwind1, and fwind2, as well as for various one-dimensional applications. [f1,f2] = freqspace(n) returns the two-dimensional frequency vectors f1 and f2 for an n-by-n matrix. For n odd, both f1 and f2 are [–n+1:2:n–1]/n. For n even, both f1 and f2 are [–n:2:n–2]/n. [f1,f2] = freqspace([m n]) returns the two-dimensional frequency vectors f1 and f2 for an m-by-n matrix. [x1,y1] = freqspace(...,'meshgrid') is equivalent to [f1,f2] = freqspace(...); [x1,y1] = meshgrid(f1,f2); f = freqspace(N) returns the one-dimensional frequency vector f assuming N evenly spaced points around the unit circle. For N even or odd, f is (0:2/N:1). For N even, freqspace therefore returns (N+2)/2 points. For N odd, it returns (N+1)/2 points. f = freqspace(N,'whole') returns N evenly spaced points around the whole unit circle. In this case, f is 0:2/N:2*(N–1)/N. Rema rk s freqspace is a function in MATLAB. See Also fsamp2, fwind1, fwind2 meshgrid in the online MATLAB Function Reference 11-64 www.MatlabSite.com freqz2 Purpose 11freqz2 Compute two-dimensional frequency response Synta x [H,f1,f2] = freqz2(h,n1,n2) [H,f1,f2] = freqz2(h,[n2 n1]) [H,f1,f2] = freqz2(h,f1,f2) [H,f1,f2] = freqz2(h) [...] = freqz2(h,...,[dx dy]) [...] = freqz2(h,...,dx) freqz2(...) Description [H,f1,f2] = freqz2(h,n1,n2) returns H, the n2-by-n1 frequency response of h, and the frequency vectors f1 (of length n1) and f2 (of length n2). h is a two-dimensional FIR filter, in the form of a computational molecule. f1 and f2 are returned as normalized frequencies in the range –1.0 to 1.0, where 1.0 corresponds to half the sampling frequency, or π radians. [H,f1,f2] = freqz2(h,[n2 n1]) returns the same result returned by [H,f1,f2] = freqz2(h,n1,n2). [H,f1,f2] = freqz2(h) uses [n2 n1] = [64 64]. [H,f1,f2] = freqz2(h,f1,f2) returns the frequency response for the FIR filter h at frequency values in f1 and f2. These frequency values must be in the range –1.0 to 1.0, where 1.0 corresponds to half the sampling frequency, or π radians. [...] = freqz2(h,...,[dx dy]) uses [dx dy] to override the intersample spacing in h. dx determines the spacing for the x-dimension and dy determines the spacing for the y-dimension. The default spacing is 0.5, which corresponds to a sampling frequency of 2.0. [...] = freqz2(h,...,dx) uses dx to determine the intersample spacing in both dimensions. With no output arguments, freqz2(...) produces a mesh plot of the two-dimensional magnitude frequency response. Cla ss Support The input matrix h can be of class uint8 or double. All other inputs to freqz2 must be of class double. All outputs are of class double. www.MatlabSite.com 11-65 freqz2 Ex a mple Use the window method to create a 16-by-16 filter, then view its frequency response using freqz2. Hd = zeros(16,16); Hd(5:12,5:12) = 1; Hd(7:10,7:10) = 0; h = fwind1(Hd,bartlett(16)); colormap(jet(64)) freqz2(h,[32 32]); axis ([–1 1 –1 1 0 1]) See Also 11-66 freqz in the Signal Processing Toolbox www.MatlabSite.com fsa m p2 Purpose 11fsamp2 Design two-dimensional FIR filter using frequency sampling Synta x h = fsamp2(Hd) h = fsamp2(f1,f2,Hd,[m n]) Description fsamp2 designs two-dimensional FIR filters based on a desired two-dimensional frequency response sampled at points on the Cartesian plane. h = fsamp2(Hd) designs a two-dimensional FIR filter with frequency response Hd, and returns the filter coefficients in matrix h. (fsamp2 returns h as a computational molecule, which is the appropriate form to use with filter2.) The filter h has a frequency response that passes through points in Hd. If Hd is m-by-n, then h is also m-by-n. Hd is a matrix containing the desired frequency response sampled at equally spaced points between –1.0 and 1.0 along the x and y frequency axes, where 1.0 corresponds to half the sampling frequency, or π radians. H d ( f 1, f 2 ) = H d ( ω 1, ω 2 ) ω 1 = π f 1, ω 2 = π f 2 For accurate results, use frequency points returned by freqspace to create Hd. (See the entry for freqspace for more information.) h = fsamp2(f1,f2,Hd,[m n]) produces an m-by-n FIR filter by matching the filter response at the points in the vectors f1 and f2. The frequency vectors f1 and f2 are in normalized frequency, where 1.0 corresponds to half the sampling frequency, or π radians. The resulting filter fits the desired response as closely as possible in the least squares sense. For best results, there must be at least m*n desired frequency points. fsamp2 issues a warning if you specify fewer than m*n points. Cla ss Support The input matrix Hd can be of class uint8 or double. All other inputs to fsamp2 must be of class double. All outputs are of class double. Ex a mple Use fsamp2 to design an approximately symmetric two-dimensional bandpass filter with passband between 0.1 and 0.5 (normalized frequency, where 1.0 corresponds to half the sampling frequency, or π radians). www.MatlabSite.com 11-67 fsa mp2 1 Create a matrix Hd that contains the desired bandpass response. Use freqspace to create the frequency range vectors f1 and f2. [f1,f2] = freqspace(21,'meshgrid'); Hd = ones(21); r = sqrt(f1.^2 + f2.^2); Hd((r<0.1)|(r>0.5)) = 0; colormap(jet(64)) mesh(f1,f2,Hd) 2 Design the filter that passes through this response. h = fsamp2(Hd); 11-68 www.MatlabSite.com fsa m p2 freqz2(h) Algorithm fsamp2 computes the filter h by taking the inverse discrete Fourier transform of the desired frequency response. If the desired frequency response is real and symmetric (zero phase), the resulting filter is also zero phase. See Also conv2, filter2, freqspace, ftrans2, fwind1, fwind2 Reference Lim, Jae S. Two-Dimensional Signal and Image Processing. Englewood Cliffs, NJ: Prentice Hall, 1990. pp. 213-217. www.MatlabSite.com 11-69 fspecia l Purpose 11fspecial Create predefined filters Synta x h = fspecial(type) h = fspecial(type,parameters) Description h = fspecial(type) creates a two-dimensional filter h of the specified type. (fspecial returns h as a computational molecule, which is the appropriate form to use with filter2.) type is a string having one of these values: • 'gaussian' for a Gaussian lowpass filter • 'sobel' for a Sobel horizontal edge-emphasizing filter • 'prewitt' for a Prewitt horizontal edge-emphasizing filter • 'laplacian' for a filter approximating the two-dimensional Laplacian operator • 'log' for a Laplacian of Gaussian filter • 'average' for an averaging filter • 'unsharp' for an unsharp contrast enhancement filter Depending on type, fspecial may take additional parameters which you can supply. These parameters all have default values. h = fspecial('gaussian',n,sigma) returns a rotationally symmetric Gaussian lowpass filter with standard deviation sigma (in pixels). n is a 1-by-2 vector specifying the number of rows and columns in h. (n can also be a scalar, in which case h is n-by-n.) If you do not specify the parameters, fspecial uses the default values of [3 3] for n and 0.5 for sigma. h = fspecial('sobel') returns this 3-by-3 horizontal edge-finding and y-derivative approximation filter: [ 1 2 1 0 0 0 –1 –2 –1 ] To find vertical edges, or for x-derivatives, use –h'. 11-70 www.MatlabSite.com fspecia l h = fspecial('prewitt') returns this 3-by-3 horizontal edge-finding and y-derivative approximation filter: [ 1 1 1 0 0 0 –1 –1 –1 ] To find vertical edges, or for x-derivatives, use –h'. h = fspecial('laplacian',alpha) returns a 3-by-3 filter approximating the two-dimensional Laplacian operator. The parameter alpha controls the shape of the Laplacian and must be in the range 0 to 1.0. fspecial uses the default value of 0.2 if you do not specify alpha. h = fspecial('log',n,sigma) returns a rotationally symmetric Laplacian of Gaussian filter with standard deviation sigma (in pixels). n is a 1-by-2 vector specifying the number of rows and columns in h. (n can also be a scalar, in which case h is n-by-n.) If you do not specify the parameters, fspecial uses the default values of [5 5] for n and 0.5 for sigma. h = fspecial('average',n) returns an averaging filter. n is a 1-by-2 vector specifying the number of rows and columns in h. (n can also be a scalar, in which case h is n-by-n.) If you do not specify n, fspecial uses the default value of [3 3]. h = fspecial('unsharp',alpha) returns a 3-by-3 unsharp contrast enhancement filter. fspecial creates the unsharp filter from the negative of the Laplacian filter with parameter alpha. alpha controls the shape of the Laplacian and must be in the range 0 to 1.0. fspecial uses the default value of 0.2 if you do not specify alpha. www.MatlabSite.com 11-71 fspecia l Ex a mple Algorithms I = imread('saturn.tif'); h = fspecial('unsharp',0.5); I2 = filter2(h,I)/255; imshow(I) figure, imshow(I2) fspecial creates Gaussian filters using: h g ( n 1, n 2 ) = e – ( n 1 + n 2 ) ⁄ ( 2σ ) 2 2 2 h g ( n 1, n 2 ) h ( n 1, n 2 ) = ---------------------------hg ∑∑ n1 n2 fspecial creates Laplacian filters using: 2 2 2 ∂ ∂ ∇ = --------- + --------2 2 ∂y ∂x α --4 4 2 1 – α ∇ ≈ ------------------ -----------(α + 1) 4 α --4 1–α ------------4 –1 1–α ------------4 α --4 1–α ------------4 α --4 fspecial creates Laplacian of Gaussian (LoG) filters using: 11-72 www.MatlabSite.com fspecia l h g ( n 1, n 2 ) = e – ( n 1 + n 2 ) ⁄ ( 2σ ) 2 2 2 2 2 2 ( n 1 + n 2 – 2σ )h g ( n 1, n 2 ) h ( n 1, n 2 ) = -----------------------------------------------------------------------6 2πσ hg ∑∑ n1 n2 fspecial creates averaging filters using: ones(n(1),n(2))/(n(1)*n(2)) fspecial creates unsharp filters using: –α 1 ------------------ α – 1 (α + 1) –α See Also α–1 α+5 α–1 –α α–1 –α conv2, edge, filter2, fsamp2, fwind1, fwind2 del2 in the online MATLAB Function Reference www.MatlabSite.com 11-73 ftra ns2 Purpose 11ftrans2 Design two-dimensional FIR filter using frequency transformation Synta x h = ftrans2(b,t) h = ftrans2(b) Description h = ftrans2(b,t) produces the two-dimensional FIR filter h that corresponds to the one-dimensional FIR filter b using the transform t. (ftrans2 returns h as a computational molecule, which is the appropriate form to use with filter2.) b must be a one-dimensional, odd-length (Type I) FIR filter such as can be returned by fir1, fir2, or remez in the Signal Processing Toolbox. The transform matrix t contains coefficients that define the frequency transformation to use. If t is m-by-n and b has length Q, then h is size ((m–1)*(Q–1)/2+1)-by-((n–1)*(Q–1)/2+1). h = ftrans2(b) uses the McClellan transform matrix t: t = [1 2 1; 2 –4 2; 1 2 1]/8; Rema rk s The transformation below defines the frequency response of the two-dimensional filter returned by ftrans2: H ( ω 1, ω 2 ) = B ( ω ) cos ω = T ( ω 1, ω 2 ) where B(ω) is the Fourier transform of the one-dimensional filter b: N B(ω) = ∑ b ( n )e – jωn n = –N and T(ω1,ω2) is the Fourier transform of the transformation matrix t: T ( ω 1, ω 2 ) = ∑ ∑ t ( n1, n2 )e – jω 1 n 1 – jω 2 n 2 e n2 n1 The returned filter h is the inverse Fourier transform of H(ω1,ω2): 1 h ( n 1, n 2 ) = -------------2( 2π ) 11-74 π π ∫–π ∫–π H ( ω1, ω2 )e jω 1 n 1 jω 2 n 2 www.MatlabSite.com e dω 1 dω 2 ftra ns2 Ex a mple Use ftrans2 to design an approximately circularly symmetric two-dimensional bandpass filter with passband between 0.1 and 0.6 (normalized frequency, where 1.0 corresponds to half the sampling frequency, or π radians). 1 Since ftrans2 transforms a one-dimensional FIR filter to create a two-dimensional filter, first design a one-dimensional FIR bandpass filter using the Signal Processing Toolbox function remez: colormap(jet(64)) b = remez(10,[0 0.05 0.15 0.55 0.65 1],[0 0 1 1 0 0]); [H,w] = freqz(b,1,128,'whole'); plot(w/pi–1,fftshift(abs(H))) 2 Use ftrans2 with the default McClellan transformation to create the desired approximately circularly symmetric filter h = ftrans2(b); www.MatlabSite.com 11-75 ftra ns2 freqz2(h) See Also conv2, filter2, fsamp2, fwind1, fwind2 Reference Lim, Jae S. Two-Dimensional Signal and Image Processing. Englewood Cliffs, NJ: Prentice Hall, 1990. pp. 218-237. 11-76 www.MatlabSite.com fw ind1 Purpose 11fwind1 Design two-dimensional FIR filter using one-dimensional window method Synta x h = fwind1(Hd,win) h = fwind1(Hd,win1,win2) h = fwind1(f1,f2,Hd,...) Description fwind1 designs two-dimensional FIR filters using the window method. fwind1 uses a one-dimensional window specification to design a two-dimensional FIR filter based on the desired frequency response Hd. fwind1 works with one-dimensional windows only; use fwind2 to work with two-dimensional windows. h = fwind1(Hd,win) designs a two-dimensional FIR filter h with frequency response Hd. (fwind1 returns h as a computational molecule, which is the appropriate form to use with filter2.) fwind1 uses the one-dimensional window win to form an approximately circularly symmetric two-dimensional window using Huang’s method. You can specify win using windows from the Signal Processing Toolbox, such as boxcar, hamming, hanning, bartlett, blackman, kaiser, or chebwin. If length(win) is n, then h is n-by-n. Hd is a matrix containing the desired frequency response sampled at equally spaced points between –1.0 and 1.0 (in normalized frequency, where 1.0 corresponds to half the sampling frequency, or π radians) along the x and y frequency axes. For accurate results, use frequency points returned by freqspace to create Hd. (See the entry for freqspace for more information.) h = fwind1(Hd,win1,win2) uses the two one-dimensional windows win1 and win2 to create a separable two-dimensional window. If length(win1) is n and length(win2) is m, then h is m-by-n. h = fwind1(f1,f2,Hd,...) lets you specify the desired frequency response Hd at arbitrary frequencies (f1 and f2) along the x and y axes. The frequency vectors f1 and f2 should be in the range –1.0 to 1.0, where 1.0 corresponds to half the sampling frequency, or π radians. The length of the window(s) controls the size of the resulting filter, as above. Ex a mple Use fwind1 to design an approximately circularly symmetric two-dimensional bandpass filter with passband between 0.1 and 0.5 (normalized frequency, where 1.0 corresponds to half the sampling frequency, or π radians). www.MatlabSite.com 11-77 fw ind1 1 Create a matrix Hd that contains the desired bandpass response. Use freqspace to create the frequency range vectors f1 and f2. [f1,f2] = freqspace(21,'meshgrid'); Hd = ones(21); r = sqrt(f1.^2 + f2.^2); Hd((r<0.1)|(r>0.5)) = 0; colormap(jet(64)) mesh(f1,f2,Hd) 2 Design the filter using a one-dimensional Hamming window. h = fwind1(Hd,hamming(21)); 11-78 www.MatlabSite.com fw ind1 freqz2(h) Algorithm fwind1 takes a one-dimensional window specification and forms an approximately circularly symmetric two-dimensional window using Huang’s method w ( n 1, n 2 ) = w ( t ) t= 2 2 n1 + n2 where w(t) is the one-dimensional window and w(n1,n2) is the resulting two-dimensional window. Given two windows, fwind1 forms a separable two-dimensional window, w ( n 1, n 2 ) = w 1 ( n 1 )w 2 ( n 2 ) fwind1 calls fwind2 with Hd and the two-dimensional window. fwind2 computes h using an inverse Fourier transform and multiplication by the two-dimensional window: jω 1 n 1 jω 2 n 2 1 π π h d ( n 1, n 2 ) = -------------2H d ( ω 1, ω 2 )e e dω 1 dω 2 – π – π ( 2π ) h ( n 1, n 2 ) = h d ( n 1, n 2 )w ( n 1, n 2 ) ∫ ∫ www.MatlabSite.com 11-79 fw ind1 See Also conv2, filter2, fsamp2, freqspace, ftrans2, fwind2 Reference Lim, Jae S. Two-Dimensional Signal and Image Processing. Englewood Cliffs, NJ: Prentice Hall, 1990. 11-80 www.MatlabSite.com fw ind2 Purpose 11fwind2 Design two-dimensional FIR filter using two-dimensional window method Synta x h = fwind2(Hd,win) h = fwind2(f1,f2,Hd,win) Description Use fwind2 to design two-dimensional FIR filters using the window method. fwind2 uses a two-dimensional window specification to design a two-dimensional FIR filter based on the desired frequency response Hd. fwind2 works with two-dimensional windows; use fwind1 to work with one-dimensional windows. h = fwind2(Hd,win) produces the two-dimensional FIR filter h using an inverse Fourier transform of the desired frequency response Hd and multiplication by the window win. Hd is a matrix containing the desired frequency response at equally spaced points in the Cartesian plane. fwind2 returns h as a computational molecule, which is the appropriate form to use with filter2. h is the same size as win. For accurate results, use frequency points returned by freqspace to create Hd. (See the entry for freqspace for more information. h = fwind2(f1,f2,Hd,win) lets you specify the desired frequency response Hd at arbitrary frequencies (f1 and f2) along the x and y axes. The frequency vectors f1 and f2 should be in the range –1.0 to 1.0, where 1.0 corresponds to half the sampling frequency, or π radians. h is the same size as win. Ex a mple Use fwind2 to design an approximately circularly symmetric two-dimensional bandpass filter with passband between 0.1 and 0.5 (normalized frequency, where 1.0 corresponds to half the sampling frequency, or π radians). www.MatlabSite.com 11-81 fw ind2 1 Create a matrix Hd that contains the desired bandpass response. Use freqspace to create the frequency range vectors f1 and f2. [f1,f2] = freqspace(21,'meshgrid'); Hd = ones(21); r = sqrt(f1.^2 + f2.^2); Hd((r<0.1)|(r>0.5)) = 0; colormap(jet(64)) mesh(f1,f2,Hd) 11-82 www.MatlabSite.com fw ind2 2 Create a two-dimensional Gaussian window using fspecial. win = fspecial('gaussian',21,2); win = win ./ max(win(:)); % Make the maximum window value be 1. mesh(win) 3 Design the filter using the window from step 2. h = fwind2(Hd,win); www.MatlabSite.com 11-83 fw ind2 freqz2(h) Algorithm fwind2 computes h using an inverse Fourier transform and multiplication by the two-dimensional window win: 1 h d ( n 1, n 2 ) = -------------2( 2π ) π π ∫–π ∫–π H d ( ω1, ω2 )e jω 1 n 1 jω 2 n 2 e dω 1 dω 2 h ( n 1, n 2 ) = h d ( n 1, n 2 )w ( n 1, n 2 ) See Also conv2, filter2, fsamp2, freqspace, ftrans2, fwind1 Reference Lim, Jae S. Two-Dimensional Signal and Image Processing. Englewood Cliffs, NJ: Prentice Hall, 1990. pp. 202-213. 11-84 www.MatlabSite.com getim a ge Purpose 11getimage Get image data from axes Synta x A = getimage(h) [x,y,A] = getimage(h) [...,A,flag] = getimage(h) [...] = getimage Description A = getimage(h) returns the first image data contained in the Handle Graphics object h. h can be a figure, axes, image, or texture-mapped surface. A is identical to the image CData; it contains the same values and is of the same class (uint8 or double) as the image CData. If h is not an image or does not contain an image or texture-mapped surface, A is empty. [x,y,A] = getimage(h) returns the image XData in x and the YData in y. XData and YData are two-element vectors that indicate the range of the x-axis and y-axis. [...,A,flag] = getimage(h) returns an integer flag that indicates the type of image h contains. This table summarizes the possible values for flag: Fla g Type of ima ge 0 Not an image; A is returned as an empty matrix 1 Intensity image with values in standard range ([0,1] for double arrays, [0,255] for uint8 arrays) 2 Indexed image 3 Intensity data, but not in standard range 4 RGB image [...] = getimage returns information for the current axes. It is equivalent to [...] = getimage(gca). Cla ss Support The output array A is of the same class as the image CData. All other inputs and outputs are of class double. www.MatlabSite.com 11-85 getima ge Ex a mple This example illustrates obtaining the image data from an image displayed directly from a file. imshow rice.tif I = getimage; 11-86 www.MatlabSite.com gra y2 ind Purpose 11gray2ind Convert an intensity image to an indexed image Synta x [X,map] = gray2ind(I,n) Description gray2ind scales, then rounds, an intensity image to produce an equivalent indexed image. [X,map] = gray2ind(I,n) converts the intensity image I to an indexed image X with colormap gray(n). If n is omitted, it defaults to 64. Cla ss Support The input image I can be of class uint8 or double. The class of the output image X is uint8 if the colormap length is less than or equal to 256. If the colormap length is greater than 256, X is of class double. See Also ind2gray www.MatlabSite.com 11-87 gra yslice Purpose 11grayslice Create indexed image from intensity image, using multilevel thresholding Synta x X = grayslice(I,n) X = grayslice(I,v) Description X = grayslice(I,n) thresholds the intensity image I using cutoff values n–1 1 2 ---, ---, …, ------------- , returning an indexed image in X. n n n X = grayslice(I,v), where v is a vector of values between 0 and 1, thresholds I using the values of v, returning an indexed image in X. You can view the thresholded image using imshow(X,map) with a colormap of appropriate length. Cla ss Support The input image I can be of class uint8 or double. Note that the threshold values are always between 0 and 1, even if I is of class uint8. In this case, each threshold value is multiplied by 255 to determine the actual threshold to use. The class of the output image X depends on the number of threshold values, as specified by n or length(v). If the number of threshold values is less than 256, then X is of class uint8, and the values in X range from 0 to n or length(v). If the number of threshold values is 256 or greater, X is of class double, and the values in X range from 1 to n+1 or length(v)+1. Ex a mple See Also 11-88 I = imread('ngc4024m.tif'); X = grayslice(I,16); imshow(I) figure, imshow(X,jet(16)) gray2ind www.MatlabSite.com histeq Purpose Synta x 11histeq Enhance contrast using histogram equalization J = histeq(I,hgram) J = histeq(I,n) [J,T] = histeq(I,...) newmap = histeq(X,map,hgram) newmap = histeq(X,map) [newmap,T] = histeq(X,...) Description histeq enhances the contrast of images by transforming the values in an intensity image, or the values in the colormap of an indexed image, so that the histogram of the output image approximately matches a specified histogram. J = histeq(I,hgram) transforms the intensity image I so that the histogram of the output intensity image J with length(hgram) bins approximately matches hgram. The vector hgram should contain integer counts for equally spaced bins with intensity values from 0 to 1.0. histeq automatically scales hgram so that sum(hgram) = prod(size(I)). The histogram of J will better match hgram when length(hgram) is much smaller than the number of discrete levels in I. J = histeq(I,n) transforms the intensity image I, returning in J an intensity image with n discrete gray levels. A roughly equal number of pixels is mapped to each of the n levels in J, so that the histogram of J is approximately flat. (The histogram of J is flatter when n is much smaller than the number of discrete levels in I.) The default value for n is 64. [J,T] = histeq(I,...) returns the gray scale transformation that maps gray levels in the intensity image I to gray levels in J. newmap = histeq(X,map,hgram) transforms the colormap associated with the indexed image X so that the histogram of the gray component of the indexed image (X,newmap) approximately matches hgram. histeq returns the transformed colormap in newmap. length(hgram) must be the same as size(map,1). newmap = histeq(X,map) transforms the values in the colormap so that the histogram of the gray component of the indexed image X is approximately flat. It returns the transformed colormap in newmap. www.MatlabSite.com 11-89 histeq [newmap,T] = histeq(X,...) returns the grayscale transformation T that maps the gray component of map to the gray component of newmap. Cla ss Support For syntaxes that include an intensity image I as input, I can be of class uint8 or double, and the output image J has the same class as I. For syntaxes that include an indexed image X as input, X can be of class uint8 or double; the output colormap is always of class double. Also, the optional output T (the gray level transform) is always of class double. Ex a mples Enhance the contrast of an intensity image using histogram equalization. I = imread('tire.tif'); J = histeq(I); imshow(I) figure, imshow(J) Display the resulting histograms. imhist(I,64) figure; imhist(J,64) 11-90 www.MatlabSite.com histeq Algorithm When you supply a desired histogram hgram, histeq chooses the grayscale transformation T to minimize c1 ( T ( k ) ) – c0 ( k ) where c0 is the cumulative histogram of A, c1 is the cumulative sum of hgram for all intensities k. This minimization is subject to the constraints that T must be monotonic and c1(T(a)) cannot overshoot c0(a) by more than half the distance between the histogram counts at a. histeq uses this transformation to map the gray levels in X (or the colormap) to their new values b = T (a) If you do not specify hgram, histeq creates a flat hgram hgram = ones(1,n)*prod(size(A))/n; and then applies the previous algorithm. See Also brighten, imadjust, imhist www.MatlabSite.com 11-91 hsv2 rgb Purpose 11hsv2rgb Convert hue-saturation-value (HSV) values to RGB color space Synta x rgbmap = hsv2rgb(hsvmap) RGB = hsv2rgb(HSV) Description rgbmap = hsv2rgb(hsvmap) converts the HSV values in hsvmap to RGB color space. hsvmap is an m-by-3 matrix that contains hue, saturation, and value components as its three columns, and rgbmap is returned as an m-by-3 matrix that represents the same set of colors as red, green, and blue values. Both rgbmap and hsvmap contain values in the range 0 to 1.0. RGB = hsv2rgb(HSV) converts the HSV image to the equivalent RGB image. HSV is an m-by-n-by-3 image array whose three planes contain the hue, saturation, and value components for the image. RGB is returned as an m-by-n-by-3 image array whose three planes contain the red, green, and blue components for the image. Cla ss Support The input array to hsv2rgb must be of class double. The output array is of class double. Rema rk s hsv2rgb is a function in MATLAB. See Also rgb2hsv, rgbplot colormap in the online MATLAB Function Reference 11-92 www.MatlabSite.com idct2 Purpose Synta x Description 11idct2 Compute two-dimensional inverse discrete cosine transform B = idct2(A) B = idct2(A,m,n) B = idct2(A,[m n]) B = idct2(A) returns the two-dimensional inverse discrete cosine transform of A. B = idct2(A,m,n) or B = idct2(A,[m n]) pads A with zeros to size m-by-n before transforming. If [m n] < size(A), idct2 crops A before transforming. For any A, idct2(dct2(A)) equals A to within roundoff error. Cla ss Support The input matrix A can be of class uint8 or double. The output matrix B is of class double. Algorithm idct2 computes the two-dimensional inverse DCT using M–1 N–1 A mn = ∑ ∑ p=0 q=0 π ( 2m + 1 ) p π ( 2n + 1 )q α p α q B pq cos ------------------------------- cos ----------------------------- , 2M 2N  1 ⁄ M, p = 0 αp =   2 ⁄ M, 1 ≤ p ≤ M – 1 0≤m≤M–1 0≤n≤N–1 1 ⁄ N , q = 0 αq =   2 ⁄ N, 1 ≤ q ≤ N – 1 See Also dct2, dctmtx, fft2, ifft2 References Jain, Anil K. Fundamentals of Digital Image Processing. Englewood Cliffs, NJ: Prentice Hall, 1989. pp. 150-153. Pennebaker, William B., and Joan L. Mitchell. JPEG: Still Image Data Compression Standard. New York: Van Nostrand Reinhold, 1993. 11 www.MatlabSite.com 11-93 ifft2 Purpose 11ifft2 Compute two-dimensional inverse fast Fourier transform Synta x B = ifft2(A) B = ifft2(A,m,n) Description B = ifft2(A) returns the two-dimensional inverse fast Fourier transform of matrix A. If A is a vector, B has the same orientation as A. B = ifft(A,m,n) pads matrix A with zeros to size m-by-n. If [m n] < size(A), ifft2 crops A before transforming. For any A, ifft2(fft2(A)) equals A to within roundoff error. If A is real, ifft2(fft2(A)) may have small imaginary parts. Cla ss Support The input matrix A can be of class uint8 or double. The output matrix B is of class double. Rema rk s ifft2 is a function in MATLAB. Algorithm The algorithm for ifft2(A) is the same as the algorithm for fft2(A), except for a sign change and scale factors of [m,n] = size(A). Like fft2, the execution time is fastest when m and n are powers of 2 and slowest when they are large prime numbers. See Also fft2, fftshift, idct2 dftmtx, filter, freqz, specplot, spectrum in the Signal Processing Toolbox User’s Guide fft, ifft in the online MATLAB Function Reference 11-94 www.MatlabSite.com ifftn Purpose 11ifftn Compute N-dimensional inverse fast Fourier transform Synta x B = ifftn(A) B = ifftn(A,siz) Description B = ifftn(A) performs the N-dimensional inverse fast Fourier transform. The result B is the same size as A. B = ifftn(A,siz) pads A with zeros (or truncates A) to create an N-dimensional array of size siz before doing the inverse transform. The size of the result is siz. For any A, ifftn(fftn(A)) equals A within roundoff error. If A is real, ifftn(fftn(A)) may have small imaginary parts. Cla ss Support The input matrix A can be of class uint8 or double. The output matrix B is of class double. Rema rk s ifftn is a function in MATLAB. Algorithm ifftn(A) is equivalent to B = A; for p = 1:length(size(A)) B = ifft(B,[],p); end This code computes the one-dimensional inverse fast Fourier transform along each dimension of A. The time required to compute ifftn(A) depends strongly on the number of prime factors of the dimensions of A. It is fastest when all of the dimensions are powers of 2. See Also fft2, fftn, ifft2 www.MatlabSite.com 11-95 im2 bw Purpose 11im2bw Convert an image to a binary image, based on threshold Synta x BW = im2bw(I,level) BW = im2bw(X,map,level) BW = im2bw(RGB,level) Description im2bw produces binary images from indexed, intensity, or RGB images. To do this, it converts the input image to grayscale format (if it is not already an intensity image), and then converts this grayscale image to binary by thresholding. The output binary image BW has values of 0 (black) for all pixels in the input image with luminance less than level and 1 (white) for all other pixels. (Note that you specify level in the range [0,1], regardless of the class of the input image.) BW = im2bw(I,level) converts the intensity image I to black and white. BW = im2bw(X,map,level) converts the indexed image X with colormap map to black and white. BW = im2bw(RGB,level) converts the RGB image RGB to black and white. Cla ss Support Ex a mple See Also 11-96 The input image can be of class uint8 or double. The output image BW is of class uint8. load trees BW = im2bw(X,map,0.4); imshow(X,map) figure, imshow(BW) ind2gray, rgb2gray www.MatlabSite.com im 2 col Purpose 11im2col Rearrange image blocks into columns Synta x B = im2col(A,[m n],block_type) B = im2col(A,[m n]) B = im2col(A,'indexed',...) Description im2col rearranges image blocks into columns. block_type is a string that can have one of these values: • 'distinct' for m-by-n distinct blocks • 'sliding' for m-by-n sliding blocks (default) B = im2col(A,[m n],'distinct') rearranges each distinct m-by-n block in the image A into a column of B. im2col pads A with zeros, if necessary, so its size is an integer multiple of m-by-n. If A = [A11 A12;A21 A22], where each Aij is m-by-n, then B = [A11(:) A12(:) A21(:) A22(:)]. B = im2col(A,[m n],'sliding') converts each sliding m-by-n block of A into a column of B, with no zero padding. B has m*n rows and will contain as many columns as there are m-by-n neighborhoods of A. If the size of A is [mm nn], then the size of B is (m*n)-by-((mm–m+1)*(nn–n+1)). B = im2col(A,[m n]) uses the default block_type of 'sliding'. For the sliding block case, each column of B contains the neighborhoods of A reshaped as nhood(:) where nhood is a matrix containing an m-by-n neighborhood of A. im2col orders the columns of B so that they can be reshaped to form a matrix in the normal way. For example, suppose you use a function, such as sum(B), that returns a scalar for each column of B. You can directly store the result in a matrix of size (mm–m+1)-by-(nn–n+1), using these calls: B = im2col(A,[m n],'sliding'); C = reshape(sum(B),mm–m+1,nn–n+1); B = im2col(A,'indexed',...) processes A as an indexed image, padding with zeros if the class of A is uint8, or ones if the class of A is double. Cla ss Support The input image A can be of class uint8 or double. The output matrix B is of the same class as the input image. See Also blkproc, col2im, colfilt, nlfilter www.MatlabSite.com 11-97 im2 double Purpose 11im2double Convert image array to double precision Synta x I2 = im2double(I1) RGB2 = im2double(RGB1) BW2 = im2double(BW1) X2 = im2double(X1,'indexed') Description im2double takes an image as input, and returns an image of class double. If the input image is of class double, the output image is identical to it. If the input image is of class uint8, im2double returns the equivalent image of class double, rescaling or offsetting the data as necessary. I2 = im2double(I1) converts the intensity image I1 to double precision, rescaling the data if necessary. RGB2 = im2double(RGB1) converts the truecolor image RGB1 to double precision, rescaling the data if necessary. BW2 = im2double(BW1) converts the binary image BW1 to double precision. X2 = im2double(X1,'indexed') converts the indexed image X1 to double precision, offsetting the data if necessary. See Also 11-98 double, im2uint8, uint8 www.MatlabSite.com im 2 uint8 Purpose 11im2uint8 Convert image array to eight-bit unsigned integers Synta x I2 = im2uint8(I1) RGB2 = im2uint8(RGB1) BW2 = im2uint8(BW1) X2 = im2uint8(X1,'indexed') Description im2uint8 takes an image as input, and returns an image of class uint8. If the input image is of class uint8, the output image is identical to it. If the input image is of class double, im2uint8 returns the equivalent image of class uint8, rescaling or offsetting the data as necessary. I2 = im2uint8(I1) converts the intensity image I1 to uint8, rescaling the data if necessary. RGB2 = im2uint8(RGB1) converts the truecolor image RGB1 to uint8, rescaling the data if necessary. BW2 = im2uint8(BW1) converts the binary image BW1 to uint8. X2 = im2uint8(X1,'indexed') converts the indexed image X1 to uint8, offsetting the data if necessary. Note that max(X1(:)) must be 256 or less, or else X1 cannot be converted to uint8. See Also double, im2double, uint8 www.MatlabSite.com 11-99 ima djust Purpose 11imadjust Adjust image intensity values or colormap Synta x J = imadjust(I,[low high],[bottom top],gamma) newmap = imadjust(map,[low high],[bottom top],gamma) Description J = imadjust(I,[low high],[bottom top],gamma) transforms the values in the intensity image I to values in J by mapping values between low and high to values between bottom and top. Values below low and above high are clipped; that is, values below low map to bottom, and those above high map to top. You can use an emptry matrix ([]) for [low high] or for [bottom top] to specify the default of [0 1]. gamma specifies the shape of the curve describing the relationship between the values in I and J. If gamma is less than 1, the mapping is weighted toward higher (brighter) output values. If gamma is greater than than 1, the mapping is weighted toward lower (darker) output values. If you omit the argument, gamma defaults to 1 (linear mapping). newmap = imadjust(map,[low high],[bottom top],gamma) transforms the colormap associated with an indexed image. If [low high] and [bottom top] are both 2-by-3, and gamma is a 1-by-3 vector, imadjust rescales the red, green, and blue components separately. The rescaled colormap, newmap, is the same size as map. RGB2 = imadjust(RGB1,...) performs the adjustment on each image plane (red, green, and blue) of the RGB image RGB1. As with the colormap adjustment, you can use different parameter values for each plane by specifying [low high] and [bottom top] as 2-by-3 matrices, and gamma as a 1-by-3 vector. Cla ss Support For syntaxes that include an input image (rather than a colormap), the input image can be of class uint8 or double. The output image has the same class as the input image. Rema rk s If top < bottom, the output image is reversed (i.e., as in a negative). 11-100 www.MatlabSite.com im a djust Ex a mple See Also I = imread('pout.tif'); J = imadjust(I,[0.3 0.7],[]); imshow(I) figure, imshow(J) brighten, histeq www.MatlabSite.com 11-101 ima pprox Purpose 11imapprox Approximate indexed image by one with fewer colors Synta x [Y,newmap] = imapprox(X,map,n) [Y,newmap] = imapprox(X,map,tol) Y = imapprox(X,map,newmap) [...] = imapprox(...,dither_option) Description [Y,newmap] = imapprox(X,map,n) approximates the colors in the indexed image X and associated colormap map by using minimum variance quantization. imapprox returns indexed image Y with colormap newmap, which has at most n colors. [Y,newmap] = imapprox(X,map,tol) approximates the colors in X and map through uniform quantization. newmap contains at most (floor(1/tol)+1)^3 colors. tol must be between 0 and 1.0. Y = imapprox(X,map,newmap) approximates the colors in map by using colormap mapping to find the colors in newmap that best match the colors in map. Y = imapprox(...,dither_option) enables or disables dithering. dither_option is a string that can have one of these values: • 'dither' dithers, if necessary, to achieve better color resolution at the expense of spatial resolution (default). • 'nodither' maps each color in the original image to the closest color in the new map. No dithering is performed. Cla ss Support The input image X can be of class uint8 or double. The output image Y is of class uint8 if the length of newmap is less than or equal to 256. If the length of newmap is greater than 256, X is of class double. Algorithm imapprox uses rgb2ind to create a new colormap that uses fewer colors. See Also cmunique, dither, rgb2ind 11-102 www.MatlabSite.com im contour Purpose 11imcontour Create a contour plot of image data Synta x imcontour(I,n) imcontour(I,v) imcontour(x,y,...) imcontour(...,LineSpec) [C,h] = imcontour(...) Description imcontour(I,n) draws a contour plot of the intensity image I, automatically setting up the axes so their orientation and aspect ratio match the image. n is the number of equally spaced contour levels in the plot; if you omit the argument, the number of levels and the values of the levels are chosen automatically. imcontour(I,v) draws a contour plot of I with contour lines at the data values specified in vector v. The number of contour levels is equal to length(v). imcontour(x,y,...) uses the vectors x and y to specify the x- and y-axis limits. imcontour(...,LineSpec) draws the contours using the line type and color specified by LineSpec. Marker symbols are ignored. [C,h] = imcontour(...) returns the contour matrix C and a vector of handles to the objects in the plot. (The objects are actually patches, and the lines are the edges of the patches.) You can use the clabel function with the contour matrix C to add contour labels to the plot. Cla ss Support The input image can be of class uint8 or double. www.MatlabSite.com 11-103 imcontour Ex a mple See Also 11-104 I = imread('ic.tif'); imcontour(I,3) clabel, contour, LineSpec in the online MATLAB Function Reference www.MatlabSite.com im crop Purpose Synta x 11imcrop Crop an image I2 = imcrop(I) X2 = imcrop(X,map) RGB2 = imcrop(RGB) I2 = imcrop(I,rect) X2 = imcrop(X,map,rect) RGB2 = imcrop(RGB,rect) [...] = imcrop(x,y,...) [A,rect] = imcrop(...) [x,y,A,rect] = imcrop(...) Description imcrop crops an image to a specified rectangle. In the syntaxes below, imcrop displays the input image and waits for you to specify the crop rectangle with the mouse: I2 = imcrop(I) X2 = imcrop(X,map) RGB2 = imcrop(RGB) If you omit the input arguments, imcrop operates on the image in the current axes. To specify the rectangle: • For a single-button mouse, press the mouse button and drag to define the crop rectangle. Finish by releasing the mouse button. • For a 2- or 3-button mouse, press the left mouse button and drag to define the crop rectangle. Finish by releasing the mouse button. If you hold down the Shift key while dragging, or if you press the right mouse button on a 2- or 3-button mouse, imcrop constrains the bounding rectangle to be a square. When you release the mouse button, imcrop returns the cropped image in the supplied output argument. If you do not supply an output argument, imcrop displays the output image in a new figure. www.MatlabSite.com 11-105 imcrop You can also specify the cropping rectangle noninteractively, using these syntaxes: I2 = imcrop(I,rect) X2 = imcrop(X,map,rect) RGB2 = imcrop(RGB,rect) rect is a 4-element vector with the form [xmin ymin width height]; these values are specified in spatial coordinates. To specify a nondefault spatial coordinate system for the input image, precede the other input arguments with two 2-element vectors specifying the XData and YData: [...] = imcrop(x,y,...) If you supply additional output arguments, imcrop returns information about the selected rectangle and the coordinate system of the input image: [A,rect] = imcrop(...) [x,y,A,rect] = imcrop(...) A is the output image. x and y are the XData and YData of the input image. Cla ss Support The input image A can be of class uint8 or double. The output image B is of the same class as A. rect is always of class double. Rema rk s Because rect is specified in terms of spatial coordinates, the width and height elements of rect do not always correspond exactly with the size of the output image. For example, suppose rect is [20 20 40 30], using the default spatial coordinate system. The upper-left corner of the specified rectangle is the center of the pixel (20,20) and the lower-right corner is the center of the pixel (50,60). The resulting output image is 31-by-41, not 30-by-40, because the output image includes all pixels in the input image that are completely or partially enclosed by the rectangle. 11-106 www.MatlabSite.com im crop Ex a mple See Also I = imread('ic.tif'); I2 = imcrop(I,[60 40 100 90]); imshow(I) figure, imshow(I2) zoom www.MatlabSite.com 11-107 imfea ture Purpose 11imfeature Compute feature measurements for image regions Synta x stats = imfeature(L,measurements) stats = imfeature(L,measurements,n) Description stats = imfeature(L,measurements) computes a set of measurements for each labeled region in the label matrix L. Positive integer elements of L correspond to different regions. For example, the set of elements of L equal to 1 corresponds to region 1; the set of elements of L equal to 2 corresponds to region 2; and so on. stats is a structure array of length max(L(:)). The fields of the structure array denote different measurements for each region, as specified by measurements. measurements can be a comma-separated list of strings, a cell array containing strings, the single string 'all', or the single string 'basic'. The set of valid measurement strings includes: 'Area' 'Image' 'EulerNumber' 'Centroid' 'FilledImage' 'Extrema' 'BoundingBox' 'FilledArea' 'EquivDiameter' 'MajorAxisLength' 'ConvexHull' 'Solidity' 'MinorAxisLength' 'ConvexImage' 'Extent' 'Eccentricity' 'ConvexArea' 'PixelList' 'Orientation' Measurement strings are case insensitive and can be abbreviated. If measurements is the string 'all', then all of the above measurements are computed. If measurements is not specified or if it is the string 'basic', then these measurements are computed: 'Area', 'Centroid', and 'BoundingBox'. stats = imfeature(L,measurements,n) specifies the type of connectivity used in computing the 'FilledImage', 'FilledArea', and 'EulerNumber' measurements. n can have a value of either 4 or 8, where 4 specifies 4-connected objects and 8 specifies 8-connected objects; if the argument is omitted, it defaults to 8. 11-108 www.MatlabSite.com im fea ture Definitions 'Area' – Scalar; the actual number of pixels in the region. (This value may differ slightly from the value returned by bwarea, which weights different patterns of pixels differently.) 'Centroid' – 1-by-2 vector; the x- and y-coordinates of the center of mass of the region. 'BoundingBox' – 1-by-4 vector; the smallest rectangle that can contain the region. The format of the vector is [x y width height], where x and y are the x- and y-coordinates of the upper-left corner of the rectangle, and width and height are the width and height of the rectangle. Note that x and y are always noninteger values, because they are the spatial coordinates for the upper-left corner of a pixel in the image; for example, if this pixel is the third pixel in the fifth row of the image, then x = 2.5 and y = 4.5. This figure illustrates the centroid and bounding box. The region consists of the white pixels; the green box is the bounding box, and the red dot is the centroid: 'MajorAxisLength' – Scalar; the length (in pixels) of the major axis of the ellipse that has the same second-moments as the region. 'MinorAxisLength' – Scalar; the length (in pixels) of the minor axis of the ellipse that has the same second-moments as the region. 'Eccentricity' – Scalar; the eccentricity of the ellipse that has the same second-moments as the region. The eccentricity is the ratio of the distance between the foci of the ellipse and its major axis length. The value is between 0 and 1. (0 and 1 are degenerate cases; an ellipse whose eccentricity is 0 is actually a circle, while an ellipse whose eccentricity is 1 is a line segment.) 'Orientation' – Scalar; the angle (in degrees) between the x-axis and the major axis of the ellipse that has the same second-moments as the region. This figure illustrates the axes and orientation of the ellipse. The left side of the figure shows an image region and its corresponding ellipse. The right side www.MatlabSite.com 11-109 imfea ture shows the same ellipse, with features indicated graphically; the solid blue lines are the axes, the red dots are the foci, and the orientation is the angle between the horizontal dotted line and the major axis: 'Image' – Binary image (uint8) of the same size as the bounding box of the region; the on pixels correspond to the region, and all other pixels are off. 'FilledImage' – Binary image (uint8) of the same size as the bounding box of the region; the on pixels correspond to the region, with all holes filled in. 'FilledArea' – Scalar; the number of on pixels in FilledImage. This figure illustrates ‘Image' and ‘FilledImage': Original image, containing a single region 'Image' 'FilledImage' 'ConvexHull' – p-by-2 matrix; the smallest convex polygon that can contain the region. Each row of the matrix contains the x- and y-coordinates of one vertex of the polygon. 'ConvexImage' – Binary image (uint8); the convex hull, with all pixels within the hull filled in (i.e., set to on). (For pixels that the boundary of the hull passes 11-110 www.MatlabSite.com im fea ture through, imfeature uses the same logic as roipoly to determine whether the pixel is inside or outside the hull.) The image is the size of the bounding box of the region. 'ConvexArea' – Scalar; the number of pixels in 'ConvexImage'. 'EulerNumber' – Scalar; equal to the number of objects in the region minus the number of holes in those objects. 'Extrema' – 8-by-2 matrix; the extremal points in the region. Each row of the matrix contains the x- and y-coordinates of one of the points; the format of the vector is [top-left top-right right-top right-bottom bottom-right bottom-left left-bottom left-top]. This figure illustrates the extrema of two different regions. In the region on the left, each extremal point is distinct; in the region on the right, certain extremal points (e.g., top-left and left-top) are identical: top-left top-right top-left top-right left-top right-top left-top right-top left-bottom right-bottom left-bottom right-bottom bottom-left bottom-right bottom-left bottom-right 'EquivDiameter' – Scalar; the diameter of a circle with the same area as the region. Computed as sqrt(4*Area/pi). 'Solidity' – Scalar; the proportion of the pixels in the convex hull that are also in the region. Computed as Area/ConvexArea. 'Extent' – Scalar; the proportion of the pixels in the bounding box that are also in the region. Computed as the Area divided by area of the bounding box. 'PixelList' – p-by-2 matrix; the actual pixels in the region. Each row of the matrix contains the x- and y-coordinates of one pixel in the region. Cla ss Support The input label matrix L can be of class uint8 or double. www.MatlabSite.com 11-111 imfea ture Rema rk s The comma-separated list syntax for structure arrays is very useful when working with the output of imfeature. For example, for a field that contains a scalar, you can use a this syntax to create a vector containing the value of this field for each region in the image. For instance, if stats is a structure array with field Area, then these two expressions are equivalent: stats(1).Area, stats(2).Area, ..., stats(end).Area and: stats.Area Therefore, you can use these calls to create a vector containing the area of each region in the image: stats = imfeature(L,'Area'); allArea = [stats.Area]; allArea is a vector of the same length as the structure array stats. The function ismember is useful in conjunction with imfeature for selecting regions based on certain criteria. For example, these commands create a binary image containing only the regions in text.tif whose area is greater than 80: idx = find([stats.Area] > 80); BW2 = ismember(L,idx); Most of the measurements take very little time to compute. The exceptions are these, which may take significantly longer, depending on the number of regions in L: • • • • 'ConvexHull' 'ConvexImage' 'ConvexArea' 'FilledImage' Note that computing certain groups of measurements takes about the same amount of time as computing just one of them, because imfeature takes advantage of intermediate computations used in both computations. Therefore, it is fastest to compute all of the desired measurements in a single call to imfeature. 11-112 www.MatlabSite.com im fea ture Ex a mple BW = imread('text.tif'); L = bwlabel(BW); stats = imfeature(L,'all'); stats(23) ans = Area: Centroid: BoundingBox: MajorAxisLength: MinorAxisLength: Eccentricity: Orientation: ConvexHull: ConvexImage: ConvexArea: Image: FilledImage: FilledArea: EulerNumber: Extrema: EquivDiameter: Solidity: Extent: PixelList: See Also 89 [95.6742 192.9775] [87.5000 184.5000 16 15] 19.9127 14.2953 0.6961 9.0845 [28x2 double] [15x16 uint8 ] 205 [15x16 uint8 ] [15x16 uint8 ] 122 0 [ 8x2 double] 10.6451 0.4341 0.3708 [89x2 double] bwlabel ismember in the online MATLAB Function Reference www.MatlabSite.com 11-113 imfinfo Purpose 11imfinfo Return information about a graphics file Synta x info = imfinfo(filename,fmt) info = imfinfo(filename) Description info = imfinfo(filename,fmt) returns a structure whose fields contain information about an image in a graphics file. filename is a string that specifies the name of the graphics file, and fmt is a string that specifies the format of the file. The file must be in the current directory or in a directory on the MATLAB path. If imfinfo cannot find a file named filename, it looks for a file named filename.fmt. This table lists the possible values for fmt: Form a t File type 'bmp' Windows Bitmap (BMP) 'hdf' Hierarchical Data Format (HDF) 'jpg' or 'jpeg' Joint Photographic Experts Group (JPEG) 'pcx' Windows Paintbrush (PCX) 'tif' or 'tiff' Tagged Image File Format (TIFF) 'xwd' X Windows Dump (XWD) If filename is a TIFF or HDF file containing more than one image, info is a structure array with one element (i.e., an individual structure) for each image in the file. For example, info(3) would contain information about the third image in the file. 11-114 www.MatlabSite.com im finfo The set of fields in info depends on the individual file and its format. However, the first nine fields are always the same. This table lists these fields and describes their values: Field Va lue Filename A string containing the name of the file; if the file is not in the current directory, the string contains the full pathname of the file FileModDate A string containing the date when the file was last modified FileSize An integer indicating the size of the file in bytes Format A string containing the file format, as specified by fmt; for JPEG and TIFF files, the three-letter variant is returned FormatVersion A string or number describing the version of the format Width An integer indicating the width of the image in pixels Height An integer indicating the height of the image in pixels BitDepth An integer indicating the number of bits per pixel ColorType A string indicating the type of image; either 'truecolor' for a truecolor RGB image, 'grayscale' for a grayscale intensity image, or 'indexed' for an indexed image info = imfinfo(filename) attempts to infer the format of the file from its contents. Rema rk s imfinfo is a function in MATLAB. www.MatlabSite.com 11-115 imfinfo Ex a mple info = imfinfo('canoe.tif') info = Filename: FileModDate: FileSize: Format: FormatVersion: Width: Height: BitDepth: ColorType: FormatSignature: ByteOrder: NewSubfileType: BitsPerSample: Compression: PhotometricInterpretation: StripOffsets: SamplesPerPixel: RowsPerStrip: StripByteCounts: XResolution: YResolution: ResolutionUnit: Colormap: PlanarConfiguration: TileWidth: TileLength: TileOffsets: TileByteCounts: Orientation: FillOrder: GrayResponseUnit: MaxSampleValue: MinSampleValue: Thresholding: See Also 11-116 'canoe.tif' '25-Oct-1996 22:10:39' 69708 'tif' [] 346 207 8 'indexed' [73 73 42 0] 'little-endian' 0 8 'PackBits' 'RGB Palette' [ 9x1 double] 1 23 [ 9x1 double] 72 72 'Inch' [256x3 double] 'Chunky' [] [] [] [] 1 1 0.0100 255 0 1 imread, imwrite www.MatlabSite.com im hist Purpose 11imhist Display a histogram of image data Synta x imhist(I,n) imhist(X,map) [counts,x] = imhist(...) Description imhist(I,n) displays a histogram with n bins for the intensity image I above a grayscale colorbar of length n. If you omit the argument, imhist uses a default value of n = 256 if I is a grayscale image, or n = 2 if I is a binary image. imhist(X,map) displays a histogram for the indexed image X. This histogram shows the distribution of pixel values above a colorbar of the colormap map. The colormap must be at least as long as the largest index in X. The histogram has one bin for each entry in the colormap. [counts,x] = imhist(...) returns the histogram counts in counts and the bin locations in x so that stem(x,counts) shows the histogram. For indexed images, it returns the histogram counts for each colormap entry; the length of counts is the same as the length of the colormap. Cla ss Support Ex a mple See Also The input image can be of class uint8 or double. I = imread('pout.tif'); imhist(I) histeq hist in the online MATLAB Function Reference www.MatlabSite.com 11-117 immovie Purpose 11immovie Make a movie of a multiframe indexed image Synta x mov = immovie(X,map) Description mov = immovie(X,map) returns the movie matrix mov from the images in the multiframe indexed image X. As it creates the movie matrix, it displays the movie frames on the screen. You can play the movie using the MATLAB movie function. X comprises multiple indexed images, all having the same size and all using the colormap map. X is an m-by-n-by-1-by-k array, where k is the number of images. Cla ss Support Ex a mple See Also X can be of class uint8 or double. mov is of class double. load mri mov = immovie(D,map); montage getframe, movie in the online MATLAB Function Reference 11-118 www.MatlabSite.com im noise Purpose 11imnoise Add noise to an image Synta x J = imnoise(I,type) J = imnoise(I,type,parameters) Description J = imnoise(I,type) adds noise of type type to the intensity image I. type is a string that can have one of these values: • 'gaussian' for Gaussian white noise • 'salt & pepper' for “on and off” pixels • 'speckle' for multiplicative noise The parameters you supply depend on the type of noise. If you omit these arguments, imnoise uses default values for the parameters. J = imnoise(I,'gaussian',m,v) adds Gaussian white noise of mean m and variance v to the image I. The default is zero mean noise with 0.01 variance. J = imnoise(I,'salt & pepper',d) adds salt and pepper noise to the image I, where d is the noise density. This affects approximately d*prod(size(I)) pixels. The default is 0.05 noise density. J = imnoise(I,'speckle',v) adds multiplicative noise to the image I, using the equation J = I + n*I, where n is uniformly distributed random noise with mean 0 and variance v. The default for v is 0.04. Cla ss Support The input image I can be of class uint8 or double. The output image J is of the same class as I. www.MatlabSite.com 11-119 imnoise Ex a mple See Also 11-120 I = imread('eight.tif'); J = imnoise(I,'salt & pepper',0.02); imshow(I) figure, imshow(J) rand, randn in the online MATLAB Function Reference www.MatlabSite.com im pix el Purpose Synta x 11impixel Determine pixel color values P = impixel(I) P = impixel(X,map) P = impixel(RGB) P = impixel(I,c,r) P = impixel(X,map,c,r) P = impixel(RGB,c,r) [c,r,P] = impixel(...) P = impixel(x,y,I,xi,yi) P = impixel(x,y,X,map,xi,yi) P = impixel(x,y,RGB,xi,yi) [xi,yi,P] = impixel(x,y,...) Description impixel returns the red, green, and blue color values of specified image pixels. In the syntaxes below, impixel displays the input image and waits for you to specify the pixels with the mouse: P = impixel(I) P = impixel(X,map) P = impixel(RGB) If you omit the input arguments, impixel operates on the image in the current axes. Use normal button clicks to select pixels. Press Backspace or Delete to remove the previously selected pixel. A shift-click, right-click, or double-click adds a final pixel and ends the selection; pressing Return finishes the selection without adding a pixel. When you finish selecting pixels, impixel returns an m-by-3 matrix of RGB values in the supplied output argument. If you do not supply an output argument, impixel returns the matrix in ans. You can also specify the pixels noninteractively, using these syntaxes: P = impixel(I,c,r) P = impixel(X,map,c,r) P = impixel(RGB,c,r) www.MatlabSite.com 11-121 impix el r and c are equal-length vectors specifying the coordinates of the pixels whose RGB values are returned in P. The k-th row of P contains the RGB values for the pixel (r(k),c(k)). If you supply three output arguments, impixel returns the coordinates of the selected pixels: [c,r,P] = impixel(...) To specify a nondefault spatial coordinate system for the input image, use these syntaxes: P = impixel(x,y,I,xi,yi) P = impixel(x,y,X,map,xi,yi) P = impixel(x,y,RGB,xi,yi) x and y are 2-element vectors specifying the image XData and YData. xi and yi are equal-length vectors specifying the spatial coordinates of the pixels whose RGB values are returned in P. If you supply three output arguments, impixel returns the coordinates of the selected pixels: [xi,yi,P] = impixel(x,y,...) Cla ss Support Ex a mple The input image can be of class uint8 or double. All other inputs and outputs are of class double. RGB = imread('flowers.tif'); c = [12 146 410]; r = [104 156 129]; pixels = impixel(RGB,c,r) pixels = 61 253 237 See Also 11-122 59 240 37 101 0 44 improfile, pixval www.MatlabSite.com im profile Purpose Synta x 11improfile Compute pixel-value cross-sections along line segments c = improfile c = improfile(n) c = improfile(I,xi,yi) c = improfile(I,xi,yi,n) [cx,cy,c] = improfile(...) [cx,cy,c,xi,yi] = improfile(...) [...] = improfile(x,y,I,xi,yi) [...] = improfile(x,y,I,xi,yi,n) [...] = improfile(...,method) Description improfile computes the intensity values along a line or a multiline path in an image. improfile selects equally spaced points along the path you specify, and then uses interpolation to find the intensity value for each point. improfile works with grayscale intensity images and RGB images. If you call improfile with one of these syntaxes, it operates interactively on the image in the current axes: c = improfile c = improfile(n) n specifies the number of points to compute the intensity value for. If you do not provide this argument, improfile chooses a value for n, roughly equal to the number of pixels the path traverses. You specify the line or path using the mouse, by clicking on points in the image. Press Backspace or Delete to remove the previously selected point. A shift-click, right-click, or double-click adds a final point and ends the selection; pressing Return finishes the selection without adding a point. When you finish selecting points, improfile returns the interpolated data values in c. c is an n-by-1 vector if the input is a grayscale intensity image, or an n-by-1-by-3 array if the input is an RGB image. www.MatlabSite.com 11-123 improfile If you omit the output argument, improfile displays a plot of the computed intensity values. If the specified path consists of a single line segment, improfile creates a two-dimensional plot of intensity values versus the distance along the line segment; if the path consists of two or more line segments, improfile creates a three-dimensional plot of the intensity values versus their x- and y-coordinates. You can also specify the path noninteractively, using these syntaxes: c = improfile(I,xi,yi) c = improfile(I,xi,yi,n) xi and yi are equal-length vectors specifying the spatial coordinates of the endpoints of the line segments. You can use these syntaxes to return additional information: [cx,cy,c] = improfile(...) [cx,cy,c,xi,yi] = improfile(...) cx and cy are vectors of length n, containing the spatial coordinates of the points at which the intensity values are computed. To specify a nondefault spatial coordinate system for the input image, use these syntaxes: [...] = improfile(x,y,I,xi,yi) [...] = improfile(x,y,I,xi,yi,n) x and y are 2-element vectors specifying the image XData and YData. [...] = improfile(...,method) uses the specified interpolation method. method is a string that can have one of these values: • 'nearest' (default) uses nearest neighbor interpolation. • 'bilinear' uses bilinear interpolation. • 'bicubic' uses bicubic interpolation. If you omit the method argument, improfile uses the default method of 'nearest'. Cla ss Support 11-124 The input image can be of class uint8 or double. All other inputs and outputs are of class double. www.MatlabSite.com im profile Ex a mple See Also I = imread('alumgrns.tif'); x = [35 338 346 103]; y = [253 250 17 148]; improfile(I,x,y), grid on impixel, pixval interp2 in the online MATLAB Function Reference www.MatlabSite.com 11-125 imrea d Purpose 11imread Read images from graphics files Synta x A = imread(filename,fmt) [X,map] = imread(filename,fmt) [...] = imread(filename) [...] = imread(...,idx) (TIFF only) [...] = imread(...,ref) (HDF only) Description A = imread(filename,fmt) reads the image in filename into A, whose class is uint8. If the file contains a grayscale intensity image, A is a two-dimensional array. If the file contains a truecolor (RGB) image, A is a three-dimensional (m-by-n-by-3) array. filename is a string that specifies the name of the graphics file, and fmt is a string that specifies the format of the file. The file must be in the current directory or in a directory in the MATLAB path, or you can specify the full pathname for a file located anywhere on your system. If imread cannot find a file named filename, it looks for a file named filename.fmt. This table lists the possible values for fmt: Form a t File type 'bmp' Windows Bitmap (BMP) 'hdf' Hierarchical Data Format (HDF) 'jpg' or 'jpeg' Joint Photographic Experts Group (JPEG) 'pcx' Windows Paintbrush (PCX) 'tif' or 'tiff' Tagged Image File Format (TIFF) 'xwd' X Windows Dump (XWD) [X,map] = imread(filename,fmt) reads the indexed image in filename into X and its associated colormap into map. X is of class uint8, and map is of class double. The colormap values are rescaled to the range [0,1]. [...] = imread(filename) attempts to infer the format of the file from its content. 11-126 www.MatlabSite.com im rea d [...] = imread(...,idx) reads in one image from a multi-image TIFF file. idx is an integer value that specifies the order in which the image appears in the file. For example, if idx is 3, imread reads the third image in the file. If you omit this argument, imread reads the first image in the file. [...] = imread(...,ref) reads in one image from a multi-image HDF file. ref is an integer value that specifies the reference number used to identify the image. For example, if ref is 12, imread reads the image whose reference number is 12. (Note that in an HDF file the reference numbers do not necessarily correspond to the order of the images in the file.) If you omit this argument, imread reads the first image in the file. This table summarizes the types of images that imread can read: Forma t Va ria nts BMP 1-bit, 4-bit, 8-bit, and 24-bit uncompressed images; 4-bit and 8-bit run-length encoded (RLE) images HDF 8-bit raster image datasets, with or without associated colormap; 24-bit raster image datasets JPEG Any baseline JPEG image; JPEG images with some commonly used extensions PCX 1-bit, 8-bit, and 24-bit images TIFF Any baseline TIFF image, including 1-bit, 8-bit, and 24-bit uncompressed images; 1-bit, 8-bit, and 24-bit images with packbit compression; 1-bit images with CCITT compression XWD 1-bit and 8-bit ZPixmaps; XYBitmaps; 1-bit XYPixmaps Rema rk s imread is a function in MATLAB. Ex a mples This example reads the sixth image in a TIFF file: [X,map] = imread('flowers.tif',6); www.MatlabSite.com 11-127 imrea d This example reads the fourth image in an HDF file: info = imfinfo('skull.hdf'); [X,map] = imread('skull.hdf',info(4).Reference); See Also imfinfo, imwrite fread in the online MATLAB Function Reference 11-128 www.MatlabSite.com im resize Purpose Synta x 11imresize Resize an image B = imresize(A,m,method) B = imresize(A,[mrows ncols],method) B = imresize(...,method,n) B = imresize(...,method,h) Description imresize resizes an image of any type using the specified interpolation method. method is a string that can have one of these values: • 'nearest' (default) uses nearest neighbor interpolation. • 'bilinear' uses bilinear interpolation. • 'bicubic' uses bicubic interpolation. If you omit the method argument, imresize uses the default method of 'nearest'. B = imresize(A,m,method) returns an image that is m times the size of A. If m is between 0 and 1.0, B is smaller than A. If m is greater than 1.0, B is larger than A. B = imresize(A,[mrows ncols],method) returns an image of size [mrows ncols]. If the specified size does not produce the same aspect ratio as the input image has, the output image is distorted. When the specified output size is smaller than the size of the input image, and method is 'bilinear' or 'bicubic', imresize applies a lowpass filter before interpolation to reduce aliasing. The default filter size is 11-by-11. You can specify a different order for the default filter using: [...] = imresize(...,method,n) n is an integer scalar specifying the size of the filter, which is n-by-n. If n is 0 (zero), imresize omits the filtering step. You can also specify your own filter h using: [...] = imresize(...,method,h) h is any two-dimensional FIR filter (such as those returned by ftrans2, fwind1, fwind2, or fsamp2). www.MatlabSite.com 11-129 imresize Cla ss Support The input image can be of class uint8 or double. The output image is of the same class as the input image. See Also interp2 in the online MATLAB Function Reference 11-130 www.MatlabSite.com im rota te Purpose 11imrotate Rotate an image Synta x B = imrotate(A,angle,method) B = imrotate(A,angle,method,'crop') Description B = imrotate(A,angle,method) rotates the image A by angle degrees in a counter-clockwise direction, using the specified interpolation method. method is a string that can have one of these values: • 'nearest' (default) uses nearest neighbor interpolation. • 'bilinear' uses bilinear interpolation. • 'bicubic' uses bicubic interpolation. If you omit the method argument, imrotate uses the default method of 'nearest'. The returned image matrix B is, in general, larger than A to include the whole rotated image. imrotate sets invalid values on the periphery of B to 0. B = imrotate(A,angle,method,'crop') rotates the image A through angle degrees and returns the central portion which is the same size as A. Cla ss Support The input image can be of class uint8 or double. The output image is of the same class as the input image. Rema rk s To rotate the image clockwise, specify a negative angle. www.MatlabSite.com 11-131 imrota te Ex a mple See Also 11-132 I = imread('ic.tif'); J = imrotate(I,–4,'bilinear','crop'); imshow(I) figure, imshow(J) imcrop, imresize www.MatlabSite.com im show Purpose Synta x 11imshow Display an image imshow(I,n) imshow(I,[low high]) imshow(BW) imshow(X,map) imshow(RGB) imshow(...,display_option) imshow(x,y,A,...) imshow filename h = imshow(...) Description imshow(I,n) displays the intensity image I with n discrete levels of gray. If you omit n, imshow uses 256 gray levels on 24-bit displays, or 64 gray levels on other systems. imshow(I,[low high]) displays I as a grayscale intensity image, specifying the data range for I. The value low (and any value less than low) displays as black, the value high (and any value greater than high) displays as white, and values in between display as intermediate shades of gray. imshow uses the default number of gray levels. If you use an empty matrix ([]) for [low high], imshow uses [min(I(:)) max(I(:))]; the minimum value in I displays as black, and the maximum value displays as white. imshow(BW) displays the binary image BW. Values of 0 display as black, and values of 1 display as white. imshow(X,map) displays the indexed image X with the colormap map. imshow(RGB) displays the truecolor image RGB. imshow(...,display_option) displays the image, calling truesize if display_option is 'truesize', or suppressing the call to truesize if display_option is 'notruesize'. Either option string can be abbreviated. If you do not supply this argument, imshow determines whether to call truesize based on the setting of the 'ImshowTruesize' preference. imshow(x,y,A,...) uses the 2-element vectors x and y to establish a nondefault spatial coordinate system, by specifying the image XData and YData. www.MatlabSite.com 11-133 imshow imshow filename displays the image stored in the graphics file filename. imshow calls imread to read the image from the file, but the image data is not stored in the MATLAB workspace. The file must be in the current directory or on the MATLAB path. h = imshow(...) returns the handle to the image object created by imshow. Cla ss Support The input image can be of class uint8 or double. Rema rk s You can use the iptsetpref function to set several toolbox preferences that modify the behavior of imshow: • 'ImshowBorder' controls whether imshow displays the image with a border around it. • 'ImshowAxesVisible' controls whether imshow displays the image with the axes box and tick labels. • 'ImshowTruesize' controls whether imshow calls the truesize function. Note that the display_option argument to imshow enables you to override the 'ImshowTruesize' preference. For more information about these preferences, see the reference entry for iptsetpref. See Also getimage, imread, iptgetpref, iptsetpref, subimage, truesize, warp image, imagesc in the online MATLAB Function Reference 11-134 www.MatlabSite.com im w rite Purpose 11imwrite Write an image to a graphics file Synta x imwrite(A,filename,fmt) imwrite(X,map,filename,fmt) imwrite(...,filename) imwrite(...,Parameter,Value,...) Description imwrite(A,filename,fmt) writes the image in A to filename. filename is a string that specifies the name of the output file, and fmt is a string that specifies the format of the file. If A is a grayscale intensity image or a truecolor (RGB) image of class uint8, imwrite writes the actual values in the array to the file. If A is of class double, imwrite rescales the values in the array before writing, using uint8(round(255*A)). This operation converts the floating-point numbers in the range [0,1] to 8-bit integers in the range [0,255]. This table lists the possible values for fmt: Form a t File type 'bmp' Windows Bitmap (BMP) 'hdf' Hierarchical Data Format (HDF) 'jpg' or 'jpeg' Joint Photographers Expert Group (JPEG) 'pcx' Windows Paintbrush (PCX) 'tif' or 'tiff' Tagged Image File Format (TIFF) 'xwd' X Windows Dump (XWD) imwrite(X,map,filename,fmt) writes the indexed image in X, and its associated colormap map, to filename. If X is of class uint8, imwrite writes the actual values in the array to the file. If X is of class double, imwrite offsets the values in the array before writing, using uint8(X–1). map must be of class double; imwrite rescales the values in map using uint8(round(255*map)). imwrite(...,filename) writes the image to filename, inferring the format to use from the filename’s extension. The extension must be one of the legal values for fmt. www.MatlabSite.com 11-135 imw rite imwrite(...,Parameter,Value,...) specifies parameters that control various characteristics of the output file. Parameters are currently supported for HDF, JPEG, and TIFF files. This table describes the available parameters for HDF files: Pa ra m eter Va lues Defa ult 'Compression' One of these strings: 'none', 'rle', 'rle' 'jpeg' 'Quality' A number between 0 and 100; parameter applies only if 'Compression' is 'jpeg'; higher numbers mean quality is better (less image degradation due to compression), but the resulting file size is larger 75 'WriteMode' One of these strings: 'overwrite', 'overwrite' 'append' This table describes the available parameters for JPEG files: 11-136 Pa ra m eter Va lues Defa ult 'Quality' A number between 0 and 100; higher numbers mean quality is better (less image degradation due to compression), but the resulting file size is larger 75 www.MatlabSite.com im w rite This table describes the available parameters for TIFF files: Pa ra m eter Va lues Defa ult 'Compression' One of these strings: 'none', 'packbits', 'ccitt'; 'ccitt' is valid for binary images only 'ccitt' for binary images; 'packbits' for all other images 'Description' Any string; fills in the ImageDescription field returned by imfinfo empty 'Resolution' A scalar value that is used for the XResolution and YResolution tags in the output file 72 This table summarizes the types of images that imwrite can write: Forma t Va ria nts BMP 8-bit uncompressed images with associated colormap; 24-bit uncompressed images HDF 8-bit raster image datasets, with or without associated colormap; 24-bit raster image datasets JPEG Baseline JPEG images1 PCX 8-bit images TIFF Baseline TIFF images, including 1-bit, 8-bit, and 24-bit uncompressed images; 1-bit, 8-bit, and 24-bit images with packbit compression; 1-bit images with CCITT compression XWD 8-bit ZPixmaps 1. Indexed images are converted to RGB before writing out JPEG files, because the JPEG format does not support indexed images. www.MatlabSite.com 11-137 imw rite Rema rk s Ex a mple See Also imwrite is a function in MATLAB. imwrite(X,map,'flowers.hdf','Compression','none',... 'WriteMode','append') imfinfo, imread fwrite in the online MATLAB Function Reference 11-138 www.MatlabSite.com ind2 gra y Purpose 11ind2gray Convert an indexed image to an intensity image Synta x I = ind2gray(X,map) Description I = ind2gray(X,map) converts the image X with colormap map to an intensity image I. ind2gray removes the hue and saturation information from the input image while retaining the luminance. Cla ss Support X can be of class uint8 or double. I is of class double. Ex a mple Algorithm load trees I = ind2gray(X,map); imshow(X,map) figure,imshow(I) ind2gray converts the colormap to NTSC coordinates using rgb2ntsc, and sets the hue and saturation components (I and Q) to zero, creating a gray colormap. ind2gray then replaces the indices in the image X with the corresponding grayscale intensity values in the gray colormap. See Also gray2ind, imshow, rgb2ntsc www.MatlabSite.com 11-139 ind2 rgb Purpose 11ind2rgb Convert an indexed image to an RGB image Synta x RGB = ind2rgb(X,map) Description RGB = ind2rgb(X,map) converts the matrix X and corresponding colormap map to RGB (truecolor) format. Cla ss Support X can be of class uint8 or double. RGB is an m-by-n-by-3 array of class double. See Also ind2gray, rgb2ind 11-140 www.MatlabSite.com iptgetpref Purpose 11iptgetpref Get Image Processing Toolbox preferences Synta x value = iptgetpref(prefname) Description value = iptgetpref(prefname) returns the value of the Image Processing Toolbox preference specified by the string prefname. Preference names are case insensitive and can be abbreviated. iptgetpref without an input argument displays the current setting of all Image Processing Toolbox preferences. Ex a mple value = iptgetpref('ImshowAxesVisible') value = off See Also imshow, iptsetpref www.MatlabSite.com 11-141 iptsetpref Purpose 11iptsetpref Set Image Processing Toolbox preferences Synta x iptsetpref(prefname,value) Description iptsetpref(prefname,value) sets the Image Processing Toolbox preference specified by the string prefname to value. The setting persists until the end of the current MATLAB session, or until you change the setting. (To make the value persist between sessions, put the command in your startup.m file.) This table describes the available preferences. Note that the preference names are case insensitive and can be abbreviated: Preference N a m e Va lues 'ImshowBorder' 'loose' (default) or 'tight' If 'ImshowBorder' is 'loose', imshow displays the image with a border between the image and the edges of the figure window, thus leaving room for axes labels, titles, etc. If 'ImshowBorder' is 'tight', imshow adjusts the figure size so that the image entirely fills the figure. (However, there may still be a border if the image is very small, or if there are other objects besides the image and its axes in the figure.) 11-142 www.MatlabSite.com iptsetpref Preference N a me Va lues 'ImshowAxesVisible' 'on' or 'off' (default) If 'ImshowAxesVisible' is 'on', imshow displays the image with the axes box and tick labels. If 'ImshowAxesVisible' is 'off', imshow displays the image without the axes box and tick labels. 'ImshowTruesize' 'auto' (default) or 'manual' If 'ImshowTruesize' is 'manual', imshow does not call truesize. If 'ImshowTruesize' is 'auto', imshow automatically decides whether to call truesize. (imshow calls truesize if there will be no other objects in the resulting figure besides the image and its axes.) You can override this setting for an individual display by specifying the display_option argument to imshow, or you can call truesize manually after displaying the image. 'TruesizeWarning' 'on' (default) or 'off' If 'TruesizeWarning' is 'on', the truesize function displays a warning if the image is too large to fit on the screen. (The entire image is still displayed, but at less than true size.) If 'TruesizeWarning' is 'off', truesize does not display the warning. Note that this preference applies even when you call truesize indirectly, such as through imshow. iptsetpref(prefname) displays the valid values for prefname. Ex a mple See Also iptsetpref('ImshowBorder','tight') imshow, iptgetpref, truesize axis in the online MATLAB Function Reference www.MatlabSite.com 11-143 ira don Purpose 11iradon Compute inverse Radon transform Synta x I = iradon(P,theta) I = iradon(P,theta,interp,filter,d,n) Description I = iradon(P,theta) reconstructs the image I from projection data in the two-dimensional array P. The columns of P are parallel beam projection data. iradon assumes that the center of rotation is the center point of the projections, which is defined as ceil(size(P,1)/2). theta describes the angles (in degrees) at which the projections were taken. It can be either a vector containing the angles or a scalar specifying D_theta, the incremental angle between projections. If theta is a vector, it must contain angles with equal spacing between them. If theta is a scalar specifying D_theta, the projections are taken at angles theta = m*D_theta, where m = 0,1,2,...,size(P,2)–1. If the input is the empty matrix ([]), D_theta defaults to 180/size(P,2). I = iradon(P,theta,interp,filter,d,n) specifies parameters to use in the inverse Radon transform. You can specify any combination of the last four arguments. iradon uses default values for any of these arguments that you omit. interp specifies the type of interpolation to use in the backprojection. The available options are listed in order of increasing accuracy and computational complexity: • 'nearest' – nearest neighbor interpolation • 'linear' – linear interpolation (default) • 'spline' – spline interpolation filter specifies the filter to use for frequency domain filtering. filter is a string that specifies any of the following standard filters: • 'Ram-Lak' – The cropped Ram-Lak or ramp filter (default). The frequency response of this filter is | f |. Because this filter is sensitive to noise in the projections, one of the filters listed below may be preferable. These filters 11-144 www.MatlabSite.com ira don multiply the Ram-Lak filter by a window that de-emphasizes high frequencies. • 'Shepp-Logan' – The Shepp-Logan filter multiplies the Ram-Lak filter by a sinc function. • 'Cosine' – The cosine filter multiplies the Ram-Lak filter by a cosine function. • 'Hamming' – The Hamming filter multiplies the Ram-Lak filter by a Hamming window. • 'Hann' – The Hann filter multiplies the Ram-Lak filter by a Hann window. d is a scalar in the range (0,1] that modifies the filter by rescaling its frequency axis. The default is 1. If d is less than 1, the filter is compressed to fit into the frequency range [0,d], in normalized frequencies; all frequencies above d are set to 0. n is a scalar that specifies the number of rows and columns in the reconstructed image. If n is not specified, the size is determined from the length of the projections: n = 2*floor(size(P,1)/(2*sqrt(2))) If you specify n, iradon reconstructs a smaller or larger portion of the image, but does not change the scaling of the data. If the projections were calculated with the radon function, the reconstructed image may not be the same size as the original image. [I,h] = iradon(...) returns the frequency response of the filter in the vector h. Cla ss Support All input arguments must be of class double. The output arguments are of class double. www.MatlabSite.com 11-145 ira don Ex a mple P = phantom(128); R = radon(P,0:179); I = iradon(R,0:179,'nearest','Hann'); imshow(P) figure, imshow(I) Algorithm iradon uses the filtered backprojection algorithm to perform the inverse Radon transform. The filter is designed directly in the frequency domain and then multiplied by the FFT of the projections. The projections are zero-padded to a power of 2 before filtering to prevent spatial domain aliasing and to speed up the FFT. See Also radon, phantom References Kak, Avinash C., and Malcolm Slaney, Principles of Computerized Tomographic Imaging. New York: IEEE Press. 11-146 www.MatlabSite.com isbw Purpose 11isbw Return true for a binary image Synta x flag = isbw(A) Description flag = isbw(A) returns 1 if A is a binary image and 0 otherwise. isbw uses these criteria to decide if A is a binary image: • If A is of class double, all values must be either 0 or 1, and the number of dimensions of A must be 2. • If A is of class uint8, its logical flag must be on, and the number of dimensions of A must be 2. Note that a four-dimensional array that contains multiple binary images returns 0, not 1. Cla ss Support A can be of class uint8 or double. See Also isind, isgray, isrgb www.MatlabSite.com 11-147 isgra y Purpose 11isgray Return true for intensity image Synta x flag = isgray(A) Description flag = isgray(A) returns 1 if A is a grayscale intensity image and 0 otherwise. isgray uses these criteria to decide if A is an intensity image: • If A is of class double, all values must be in the range [0,1], and the number of dimensions of A must be 2. • If A is of class uint8, the number of dimensions of A must be 2. Note that a four-dimensional array that contains multiple intensity images returns 0, not 1. Cla ss Support A can be of class uint8 or double. See Also isbw, isind, isrgb 11-148 www.MatlabSite.com isind Purpose 11isind Return true for an indexed image Synta x flag = isind(A) Description flag = isind(A) returns 1 if A is an indexed image and 0 otherwise. isind uses these criteria to determine if A is an indexed image: • If A is of class double, all values in A must be integers greater than or equal to 1, and the number of dimensions of A must be 2. • If A is of class uint8, its logical flag must be off, and the number of dimensions of A must be 2. Note that a four-dimensional array that contains multiple indexed images returns 0, not 1. Cla ss Support A can be of class uint8 or double. See Also isbw, isgray, isrgb www.MatlabSite.com 11-149 isrgb Purpose 11isrgb Return true for an RGB image Synta x flag = isrgb(A) Description flag = isrgb(A) returns 1 if A is an RGB truecolor image and 0 otherwise. isrgb uses these criteria to determine if A is an RGB image: • If A is of class double, all values must be in the range [0,1], and A must be m-by-n-by-3. • If A is of class uint8, A must be m-by-n-by-3. Note that a four-dimensional array that contains multiple RGB images returns 0, not 1. Cla ss Support A can be of class uint8 or double. See Also isbw, isgray, isind 11-150 www.MatlabSite.com m a k elut Purpose 11makelut Construct a lookup table for use with applylut Synta x lut = makelut(fun,n) lut = makelut(fun,n,P1,P2,...) Description lut = makelut(fun,n) returns a lookup table for use with applylut. fun is either a string containing the name of a function or an inline function object. The function should take a 2-by-2 or 3-by-3 matrix of 1’s and 0’s as input and return a scalar. n is either 2 or 3, indicating the size of the input to fun. makelut creates lut by passing all possible 2-by-2 or 3-by-3 neighborhoods to fun, one at a time, and constructing either a 16-element vector (for 2-by-2 neighborhoods) or a 512-element vector (for 3-by-3 neighborhoods). The vector consists of the output from fun for each possible neighborhood. lut = makelut(fun,n,P1,P2,...) passes the additional parameters P1,P2,..., to fun. Cla ss Support lut is returned as a vector of class double. 11 www.MatlabSite.com 11-151 ma k elut Ex a mple In this example, the function returns 1 (true) if the number of 1’s in the neighborhood is 2 or greater, and returns 0 (false) otherwise. makelut then uses the function to construct a lookup table for 2-by-2 neighborhoods. f = inline('sum(x(:)) >= 2'); lut = makelut(f,2) lut = 0 0 0 1 0 1 1 1 0 1 1 1 1 1 1 1 See Also 11-152 applylut www.MatlabSite.com m a t2 gra y Purpose 11mat2gray Convert a matrix to a grayscale intensity image Synta x I = mat2gray(A,[amin amax]) I = mat2gray(A) Description I = mat2gray(A,[amin amax]) converts the matrix A to the intensity image I. The returned matrix I contains values in the range 0 (black) to 1.0 (full intensity or white). amin and amax are the values in A that correspond to 0 and 1.0 in I. I = mat2gray(A) sets the values of amin and amax to the minimum and maximum values in A. Cla ss Support Ex a mple See Also The input array A and the output image I are of class double. I = imread('rice.tif'); J = filter2(fspecial('sobel'),I); K = mat2gray(J); imshow(I) figure, imshow(K) gray2ind www.MatlabSite.com 11-153 mea n2 Purpose 11mean2 Compute the mean of the elements of a matrix Synta x b = mean2(A) Description b = mean2(A) computes the mean of the values in A. Cla ss Support A is an array of class uint8 or double. b is a scalar of class double. Algorithm mean2 computes the mean of an array A using mean(A(:)). See Also std2 mean, std in the online MATLAB Function Reference 11-154 www.MatlabSite.com m edfilt2 Purpose 11medfilt2 Perform two-dimensional median filtering Synta x B = medfilt2(A,[m n]) B = medfilt2(A) B = medfilt2(A,'indexed',...) Description Median filtering is a nonlinear operation often used in image processing to reduce “salt and pepper” noise. Median filtering is more effective than convolution when the goal is to simultaneously reduce noise and preserve edges. B = medfilt2(A,[m n]) performs median filtering of the matrix A in two dimensions. Each output pixel contains the median value in the m-by-n neighborhood around the corresponding pixel in the input image. medfilt2 pads the image with zeros on the edges, so the median values for the points within [m n]/2 of the edges may appear distorted. B = medfilt2(A) performs median filtering of the matrix A using the default 3-by-3 neighborhood. B = medfilt2(A,'indexed',...) processes A as an indexed image, padding with zeros if the class of A is uint8, or ones if the class of A is double. Cla ss Support The input image A can be of class uint8 or double. The output image B is of the same class as A. Rema rk s If the input image A is of class uint8, all of the output values are returned as uint8 integers. If the number of pixels in the neighborhood (i.e., m*n) is even, some of the median values may not be integers. In these cases, the fractional parts are discarded. For example, suppose you call medfilt2 using 2-by-2 neighborhoods, and the input image is a uint8 array that includes this neighborhood: 1 5 4 8 medfilt2 returns an output value of 4 for this neighborhood, although the true median is 4.5. www.MatlabSite.com 11-155 medfilt2 Ex a mple This example adds salt and pepper noise to an image, then restores the image using medfilt2. I = imread('eight.tif'); J = imnoise(I,'salt & pepper',0.02); K = medfilt2(J); imshow(J) figure, imshow(K) Algorithm medfilt2 uses ordfilt2 to perform the filtering. See Also filter2, ordfilt2, wiener2 Reference Lim, Jae S. Two-Dimensional Signal and Image Processing. Englewood Cliffs, NJ: Prentice Hall, 1990. pp. 469-476. 11-156 www.MatlabSite.com m onta ge Purpose 11montage Display multiple image frames as a rectangular montage Synta x montage(I) montage(BW) montage(X,map) montage(RGB) h = montage(...) Description montage displays all of the frames of a multiframe image array in a single image object, arranging the frames so that they roughly form a square. montage(I) displays the k frames of the intensity image array I. I is m-by-n-by-1-by-k. montage(BW) displays the k frames of the binary image array BW. BW is m-by-n-by-1-by-k. montage(X,map) displays the k frames of the indexed image array X, using the colormap map for all frames. X is m-by-n-by-1-by-k. montage(RGB) displays the k frames of the truecolor image array RGB. RGB is m-by-n-by-3-by-k. h = montage(...) returns the handle to the image object. Cla ss Support The input image can be of class uint8 or double. www.MatlabSite.com 11-157 monta ge Ex a mple See Also 11-158 load mri montage(D,map) immovie www.MatlabSite.com nlfilter Purpose 11nlfilter Perform general sliding-neighborhood operations Synta x B = nlfilter(A,[m n],fun) B = nlfilter(A,[m n],fun,P1,P2,...) B = nlfilter(A,'indexed',...) Description B = nlfilter(A,[m n],fun) applies the function fun to each m-by-n sliding block of A. fun can be a string containing the name of a function, a string containing an expression, or an inline function object. fun should accept an m-by-n block as input, and return a scalar result: c = fun(x) c is the output value for the center pixel in the m-by-n block x. nlfilter calls fun for each pixel in A. nlfilter zero pads the m-by-n block at the edges, if necessary. B = nlfilter(A,[m n],fun,P1,P2,...) passes the additional parameters P1,P2,..., to fun. B = nlfilter(A,'indexed',...) processes A as an indexed image, padding with ones if A is of class double and zeros if A is of class uint8. Cla ss Support The input image A can be of any class supported by fun. The class of B depends on the class of the output from fun. Rema rk s nlfilter can take a long time to process large images. In some cases, the colfilt function can perform the same operation much faster. Ex a mple This call produces the same result as calling medfilt2 with a 3-by-3 neighborhood: B = nlfilter(A,[3 3],'median(x(:))'); See Also blkproc, colfilt www.MatlabSite.com 11-159 ntsc2 rgb Purpose 11ntsc2rgb Convert NTSC values to RGB color space Synta x rgbmap = ntsc2rgb(yiqmap) RGB = ntsc2rgb(YIQ) Description rgbmap = ntsc2rgb(yiqmap) converts the m-by-3 NTSC (television) color values in yiqmap to RGB color space. If yiqmap is m-by-3 and contains the NTSC luminance (Y) and chrominance (I and Q) color components as columns, then rgbmap is an m-by-3 matrix that contains the red, green, and blue values equivalent to those colors. Both rgbmap and yiqmap contain intensities in the range 0 to 1.0. The intensity 0 corresponds to the absence of the component, while the intensity 1.0 corresponds to full saturation of the component. RGB = ntsc2rgb(YIQ) converts the NTSC image YIQ to the equivalent truecolor image RGB. ntsc2rgb computes the RGB values from the NTSC components using: R 1.000 G = 1.000 B 1.000 0.956 – 0.272 – 1.106 0.621 Y – 0.647 I 1.703 Q Cla ss Support The input image or colormap must be of class double. The output is of class double. See Also rgb2ntsc, rgb2ind, ind2rgb, ind2gray 11-160 www.MatlabSite.com ordfilt2 Purpose 11ordfilt2 Perform two-dimensional order-statistic filtering Synta x Y = ordfilt2(X,order,domain) Y = ordfilt2(X,order,domain,S) Description Y = ordfilt2(X,order,domain) replaces each element in X by the order-th element in the sorted set of neighbors specified by the nonzero elements in domain. Y = ordfilt2(X,order,domain,S), where S is the same size as domain, uses the values of S corresponding to the nonzero values of domain as additive offsets. Cla ss Support The class of X may be double or uint8. The class of Y is the same as the class of X, unless the additive offset form of ordfilt2 is used, in which case the class of Y is double. Rema rk s domain is equivalent to the structuring element used for binary image operations. It is a matrix containing only 1’s and 0’s; the 1’s define the neighborhood for the filtering operation. For example, Y = ordfilt2(X,5,ones(3,3)) implements a 3-by-3 median filter; Y = ordfilt2(X,1,ones(3,3)) implements a 3-by-3 minimum filter; and Y = ordfilt2(X,9,ones(3,3)) implements a 3-by-3 maximum filter. Y = ordfilt2(X,1,[0 1 0; 1 0 1; 0 1 0]) replaces each element in X by the minimum of its north, east, south, and west neighbors. The syntax that includes S (the matrix of additive offsets) can be used to implement grayscale morphological operations, including grayscale dilation and erosion. See Also medfilt2 Reference Haralick, Robert M., and Linda G. Shapiro. Computer and Robot Vision, Volume I. Addison-Wesley, 1992. www.MatlabSite.com 11-161 pha ntom Purpose 11phantom Generate a head phantom image Synta x P = phantom(def,n) P = phantom(E,n) [p,E] = phantom(...) Description P = phantom(def,n) generates an image of a head phantom that can be used to test the numerical accuracy of radon and iradon or other two-dimensional reconstruction algorithms. P is a grayscale intensity image that consists of one large ellipse (representing the brain) containing several smaller ellipses (representing features in the brain). def is a string that specifies the type of head phantom to generate. Valid values are: • 'Shepp-Logan' – a test image used widely by researchers in tomography. • 'Modified Shepp-Logan' (default) – a variant of the Shepp-Logan phantom in which the contrast is improved for better visual perception. n is a scalar that specifies the number of rows and columns in P. If you omit the argument, n defaults to 256. P = phantom(E,n) generates a user-defined phantom, where each row of the matrix E specifies an ellipse in the image. E has 6 columns, with each column 11-162 www.MatlabSite.com pha ntom containing a different parameter for the ellipses. This table describes the columns of the matrix: Colum n Pa ra meter M ea ning Column 1 A Additive intensity value of the ellipse Column 2 a Length of the horizontal semi-axis of the ellipse Column 3 b Length of the vertical semi-axis of the ellipse Column 4 x0 x-coordinate of the center of the ellipse Column 5 y0 y-coordinate of the center of the ellipse Column 6 phi Angle (in degrees) between the horizontal semi-axis of the ellipse and the x-axis of the image For purposes of generating the phantom, the domains for the x- and y-axes span [–1,1]. Columns 2 through 5 must be specified in terms of this range. [P,E] = phantom(...) returns the matrix E used to generate the phantom. Cla ss Support All inputs must be of class double. All outputs are of class double. Rema rk s For any given pixel in the output image, the pixel’s value is equal to the sum of the additive intensity values of all ellipses that the pixel is a part of. If a pixel is not part of any ellipse, its value is 0. The additive intensity value A for an ellipse can be positive or negative; if it is negative, the ellipse will be darker than the surrounding pixels. Note that, depending on the values of A, some pixels may have values outside the range [0,1]. www.MatlabSite.com 11-163 pha ntom Ex a mple P = phantom('Modified Shepp-Logan',200); imshow(P) References Jain, Anil K. Fundamentals of Digital Image Processing. Englewood Cliffs, NJ: Prentice Hall, 1989. p. 439. See Also radon, iradon 11-164 www.MatlabSite.com pix va l Purpose 11pixval Display information about image pixels Synta x pixval on pixval off pixval pixval(fig,option) Purpose pixval on turns on interactive display of information about image pixels in the current figure. pixval installs a black bar at the bottom of the figure, which displays the row-column coordinates for whatever pixel the cursor is currently over, and the color information for that pixel. If the image is binary or intensity, the color information is a single intensity value. If the image is indexed or RGB, the color information is an RGB triplet. The values displayed are the actual data values, regardless of the class of the image array, or whether the data is in normal image range. If you click on the image and hold down the mouse button while you move the cursor, pixval also displays the Euclidean distance between the point you clicked on and the current cursor location. pixval draws a line between these points to indicate the distance being measured. When you release the mouse button, the line and the distance display disappear. You can move the display bar by clicking on it and dragging it to another place in the figure. pixval off turns interactive display off in the current figure. You can also turn off the display by clicking the button on the right side of the display bar. pixval toggles interactive display on or off in the current figure. pixval(fig,option) applies the pixval command to the figure specified by fig. option is string containing 'on' or 'off'. Rema rk s pixval conflicts with certain other functions, such as zoom and rotate3d, that enable you to use a mouse to interact with the figure. Use only one of these functions at a time. See Also impixel, improfile www.MatlabSite.com 11-165 qtdecom p Purpose Synta x 11qtdecomp Perform quadtree decomposition S S S S = = = = qtdecomp(I) qtdecomp(I,threshold) qtdecomp(I,threshold,mindim) qtdecomp(I,threshold,[mindim maxdim]) S = qtdecomp(I,fun) S = qtdecomp(I,fun,P1,P2,...) Description qtdecomp divides a square image into four equal-sized square blocks, and then tests each block to see if it meets some criterion of homogeneity. If a block meets the criterion, it is not divided any further. If it does not meet the criterion, it is subdivided again into four blocks, and the test criterion is applied to those blocks. This process is repeated iteratively until each block meets the criterion. The result may have blocks of several different sizes. S = qtdecomp(I) performs a quadtree decomposition on the intensity image I, and returns the quadtree structure in the sparse matrix S. If S(k,m) is nonzero, then (k,m) is the upper-left corner of a block in the decomposition, and the size of the block is given by S(k,m). By default, qtdecomp splits a block unless all elements in the block are equal. S = qtdecomp(I,threshold) splits a block if the maximum value of the block elements minus the minimum value of the block elements is greater than threshold. threshold is specified as a value between 0 and 1, even if I is of class uint8. If I is uint8, the threshold value you supply is multiplied by 255 to determine the actual threshold to use. S = qtdecomp(I,threshold,mindim) will not produce blocks smaller than mindim, even if the resulting blocks do not meet the threshold condition. S = qtdecomp(I,threshold,[mindim maxdim]) will not produce blocks smaller than mindim or larger than maxdim. Blocks larger than maxdim are split even if they meet the threshold condition. maxdim/mindim must be a power of 2. S = qtdecomp(I,fun) uses the function fun to determine whether to split a block. qtdecomp calls fun with all the current blocks of size m-by-m stacked into an m-by-m-by-k array, where k is the number of m-by-m blocks. fun should return a k-element vector, containing only 1’s and 0’s, where 1 indicates that the 11-166 www.MatlabSite.com qtdecomp corresponding block should be split, and 0 indicates it should not be split. (For example, if k(3) is 0, the third m-by-m block should not be split.) fun can be a string containing the name of a function, a string containing an expression, or an inline function. S = qtdecomp(I,fun,P1,P2,...) passes P1,P2,..., as additional arguments to fun. Cla ss Support For the syntaxes that do not include a function,the input image can be of class uint8 or double. For the syntaxes that include a function, the input image can be of any class supported by the function. The output matrix is always of class sparse. Rema rk s qtdecomp is appropriate primarily for square images whose dimensions are a power of 2, such as 128-by-128 or 512-by-512. These images can be divided until the blocks are as small as 1-by-1. If you use qtdecomp with an image whose dimensions are not a power of 2, at some point the blocks cannot be divided further. For example, if an image is 96-by-96, it can be divided into blocks of size 48-by-48, then 24-by-24, 12-by-12, 6-by-6, and finally 3-by-3. No further division beyond 3-by-3 is possible. To process this image, you must set mindim to 3 (or to 3 times a power of 2); if you are using the syntax that includes a function, the function must return 0 at the point when the block cannot be divided further. www.MatlabSite.com 11-167 qtdecom p Ex a mple I = [1 1 1 1 20 20 20 22 1 1 1 1 22 22 22 22 1 2 1 1 20 22 20 20 1 1 1 1 22 20 20 20 2 4 10 20 1 5 9 13 3 5 15 25 2 6 10 14 6 6 7 7 3 7 11 15 0 0 0 0 0 0 0 0 2 0 1 1 2 0 2 0 0 0 1 1 0 0 0 0 2 0 2 0 2 0 2 0 6 8 7 7 4 8 12 16]; S = qtdecomp(I,5); full(S) ans = 4 0 0 0 4 0 0 0 See Also 11-168 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 qtgetblk, qtsetblk www.MatlabSite.com 0 0 0 0 0 0 0 0 qtgetblk Purpose 11qtgetblk Get block values in quadtree decomposition Synta x [vals,r,c] = qtgetblk(I,S,dim) [vals,idx] = qtgetblk(I,S,dim) Description [vals,r,c] = qtgetblk(I,S,dim) returns in vals an array containing the dim-by-dim blocks in the quadtree decomposition of I. S is the sparse matrix returned by qtdecomp; it contains the quadtree structure. vals is a dim-by-dim-by-k array, where k is the number of dim-by-dim blocks in the quadtree decomposition; if there are no blocks of the specified size, all outputs are returned as empty matrices. r and c are vectors containing the row and column coordinates of the upper-left corners of the blocks. [vals,idx] = qtgetblk(I,S,dim) returns in idx a vector containing the linear indices of the upper-left corners of the blocks. Cla ss Support I can be of class uint8 or double. S is of class sparse. Rema rk s The ordering of the blocks in vals matches the columnwise order of the blocks in I. For example, if vals is 4-by-4-by-2, vals(:,:,1) contains the values from the first 4-by-4 block in I, and vals(:,:,2) contains the values from the second 4-by-4 block. www.MatlabSite.com 11-169 qtgetblk Ex a mple This example continues the qtdecomp example. [vals,r,c] = qtgetblk(I,S,4) vals(:,:,1) = 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 20 22 20 20 22 20 20 20 vals(:,:,2) = 20 20 20 22 22 22 22 22 r = 1 5 c = 1 1 See Also 11-170 qtdecomp, qtsetblk www.MatlabSite.com qtsetblk Purpose 11qtsetblk Set block values in quadtree decomposition Synta x J = qtsetblk(I,S,dim,vals) Description J = qtsetblk(I,S,dim,vals) replaces each dim-by-dim block in the quadtree decomposition of I with the corresponding dim-by-dim block in vals. S is the sparse matrix returned by qtdecomp; it contains the quadtree structure. vals is a dim-by-dim-by-k array, where k is the number of dim-by-dim blocks in the quadtree decomposition. Cla ss Support I can be of class uint8 or double. S is of class sparse. Rema rk s The ordering of the blocks in vals must match the columnwise order of the blocks in I. For example, if vals is 4-by-4-by-2, vals(:,:,1) contains the values used to replace the first 4-by-4 block in I, and vals(:,:,2) contains the values for the second 4-by-4 block. Ex a mple This example continues the qtgetblock example. newvals = cat(3,zeros(4),ones(4)); J = qtsetblk(I,S,4,newvals) J = 0 0 0 0 1 1 1 1 See Also 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 2 4 10 20 1 5 9 13 3 5 15 25 2 6 10 14 6 6 7 7 3 7 11 15 6 8 7 7 4 8 12 16 qtdecomp, qtgetblk www.MatlabSite.com 11-171 ra don Purpose 11radon Compute Radon transform Synta x R = radon(I,theta) R = radon(I,theta,n) [R,xp] = radon(...) Description The radon function computes the Radon transform, which is the projection of the image intensity along a radial line oriented at a specified angle. R = radon(I,theta) returns the Radon transform of the intensity image I for the angle theta degrees. If theta is a scalar, the result R is a column vector containing the Radon transform for theta degrees. If theta is a vector, then R is a matrix in which each column is the Radon transform for one of the angles in theta. If you omit theta, it defaults to 0:179. R = radon(I,theta,n) returns a Radon transform with the projection computed at n points. R has n rows. If you do not specify n, the number of points at which the projection is computed is: 2*ceil(norm(size(I)–floor((size(I)–1)/2)–1))+3 This number is sufficient to compute the projection at unit intervals, even along the diagonal. [R,xp] = radon(...) returns a vector xp containing the radial coordinates corresponding to each row of R. Cla ss Support I can be of class uint8 or double. All other inputs and outputs are of class double. Rema rk s The radial coordinates returned in xp are the values along the x'-axis, which is oriented at theta degrees counterclockwise from the x-axis. The origin of both axes is the center pixel of the image, which is defined as: floor((size(I)+1)/2) For example, in a 20-by-30 image, the center pixel is (10,15). 11-172 www.MatlabSite.com ra don Ex a mple iptsetpref('ImshowAxesVisible','on') I = zeros(100,100); I(25:75,25:75) = 1; theta = 0:180; [R,xp] = radon(I,theta); imshow(theta,xp,R,[]), colormap(hot), colorbar See Also iradon, phantom References Lim, Jae S. Two-Dimensional Signal and Image Processing. Englewood Cliffs, NJ: Prentice Hall, 1990. pp. 42-45. Bracewell, Ronald N. Two-Dimensional Imaging. Englewood Cliffs, NJ: Prentice Hall, 1995. pp. 505-537. www.MatlabSite.com 11-173 rgb2 gra y Purpose 11rgb2gray Convert an RGB image or colormap to grayscale Synta x I = rgb2gray(RGB) newmap = rgb2gray(map) Description rgb2gray converts RGB images to grayscale by eliminating the hue and saturation information while retaining the luminance. I = rgb2gray(RGB) converts the truecolor image RGB to the grayscale intensity image I. newmap = rgb2gray(map) returns a grayscale colormap equivalent to map. Cla ss Support If the input is an RGB image, it can be of class uint8 or double; the output image I is of the same class as the input image. If the input is a colormap, the input and output colormaps are both of class double. Algorithm rgb2gray converts the RGB values to NTSC coordinates, sets the hue and saturation components to zero, and then converts back to RGB color space. See Also 11-174 ind2gray, ntsc2rgb, rgb2ind, rgb2ntsc www.MatlabSite.com rgb2 hsv Purpose 11rgb2hsv Convert RGB values to hue-saturation-value (HSV) color space Synta x hsvmap = rgb2hsv(rgbmap) HSV = rgb2hsv(RGB) Description hsvmap = rgb2hsv(rgbmap) converts the m-by-3 RGB values in RGB to HSV color space. hsvmap is an m-by-3 matrix that contains the hue, saturation, and value components as columns that are equivalent to the colors in the RGB colormap. Both rgbmap and hsvmap are of class double and contain values in the range 0 to 1.0. HSV = rgb2hsv(RGB) converts the truecolor image RGB to the equivalent HSV image HSV. Cla ss Support If the input is an RGB image, it can be of class uint8 or double; the output image is of class double. If the input is a colormap, the input and output colormaps are both of class double. Rema rk s rgb2hsv is a function in MATLAB. See Also hsv2rgb, rgbplot colormap in the online MATLAB Function Reference www.MatlabSite.com 11-175 rgb2 ind Purpose 11rgb2ind Convert an RGB image to an indexed image Synta x [X,map] = rgb2ind(RGB) [X,map] = rgb2ind(RGB,tol) [X,map] = rgb2ind(RGB,n) X = rgb2ind(RGB,map) [...] = rgb2ind(...,dither_option) Description rgb2ind converts RGB images to indexed images using one of four different methods: direct translation, uniform quantization, minimum variance quantization, and colormap mapping. For all methods except direct translation, rgb2ind dithers the image unless you specify 'nodither' for dither_option. [X,map] = rgb2ind(RGB) converts the RGB image in the array RGB to an indexed image X with colormap map using direct translation. The resulting colormap may be very long, as it has one entry for each pixel in RGB. Do not set dither_option if you use this method. [X,map] = rgb2ind(RGB,tol) converts the RGB image to an indexed image X using uniform quantization. map contains at most (floor(1/tol)+1)^3 colors. tol must be between 0 and 1.0. [X,map] = rgb2ind(RGB,n) converts the RGB image to an indexed image X using minimum variance quantization. map contains at most n colors. X = rgb2ind(RGB,map) converts the RGB image to an indexed image X with colormap map by matching colors in RGB with the nearest color in the colormap map. [...] = rgb2ind(...,dither_option) enables or disables dithering. dither_option is a string that can have one of these values: • 'dither' (default) dithers, if necessary, to achieve better color resolution at the expense of spatial resolution. • 'nodither' maps each color in the original image to the closest color in the new map. No dithering is performed. Cla ss Support 11-176 The input image can be of class uint8 or double. The output image is of class uint8 if the length of map is less than or equal to 256, or double otherwise. www.MatlabSite.com rgb2 ind Rema rk s If you specify tol, rgb2ind uses uniform quantization to convert the image. This method involves cutting the RGB color cube into smaller cubes of length tol. For example, if you specify a tol of 0.1, the edges of the cubes are one-tenth the length of the RGB cube. The total number of small cubes is: n = (floor(1/tol)+1)^3 Each cube represents a single color in the output image. Therefore, the maximum length of the colormap is n. rgb2ind removes any colors that don’t appear in the input image, so the actual colormap may be much smaller than n. If you specify n, rgb2ind uses minimum variance quantization. This method involves cutting the RGB color cube into smaller boxes (not necessarily cubes) of different sizes, depending on how the colors are distributed in the image. If the input image actually uses fewer colors than the number you specify, the output colormap is also smaller. If you specify map, rgb2ind uses colormap mapping, which involves finding the colors in map that best match the colors in the RGB image. Ex a mple See Also RGB = imread('flowers.tif'); [X,map] = rgb2ind(RGB,128); imshow(X,map) cmunique, dither, imapprox, ind2rgb, rgb2gray www.MatlabSite.com 11-177 rgb2 ntsc Purpose 11rgb2ntsc Convert RGB values to NTSC color space Synta x yiqmap = rgb2ntsc(rgbmap) YIQ = rgb2ntsc(RGB) Description yiqmap = rgb2ntsc(rgbmap) converts the m-by-3 RGB values in rbgmap to NTSC color space. yiqmap is an m-by-3 matrix that contains the NTSC luminance (Y) and chrominance (I and Q) color components as columns that are equivalent to the colors in the RGB colormap. YIQ = rgb2ntsc(RGB) converts the truecolor image RGB to the equivalent NTSC image YIQ. rgb2ntsc defines the NTSC components using: Y 0.299 = I 0.596 Q 0.211 0.587 – 0.274 – 0.523 0.114 R – 0.322 G 0.312 B Cla ss Support If the input is an RGB image, it can be of class uint8 or double; the output image is of class double. If the input is a colormap, the input and output colormaps are both of class double. Rema rk s In the NTSC color space, the luminance is the grayscale signal used to display pictures on monochrome (black and white) televisions. The other components carry the hue and saturation information. See Also ntsc2rgb, rgb2ind, ind2rgb, ind2gray 11-178 www.MatlabSite.com rgb2 ycbcr Purpose 11rgb2ycbcr Convert RGB values to YCbCr color space Synta x ycbcrmap = rgb2ycbcr(rgbmap) YCBCR = rgb2ycbcr(RGB) Description ycbcrmap = rgb2ycbcr(rgbmap) converts the RGB values in rbgmap to the YCbCr color space. ycbcrmap is an m-by-3 matrix that contains the YCbCr luminance (Y) and chrominance (Cb and Cr) color components as columns. Each row represents the equivalent color to the corresponding row in the RGB colormap. YCBCR = rgb2ycbcr(RGB) converts the truecolor image RGB to the equivalent image in the YCbCr color space. Cla ss Support If the input is an RGB image, it can be of class uint8 or double; the output image is of the same class as the input image. If the input is a colormap, the input and output colormaps are both of class double. See Also ntsc2rgb, rgb2ntsc, ycbcr2rgb www.MatlabSite.com 11-179 rgbplot Purpose 11rgbplot Plot colormap Synta x rgbplot(map) Description rgbplot(map) plots the three columns of map, where map is an m-by-3 colormap matrix. rgbplot draws the first column in red, the second in green, and the third in blue. Ex a mple See Also 11-180 rgbplot(jet) colormap in the online MATLAB Function Reference www.MatlabSite.com roicolor Purpose 11roicolor Select region of interest, based on color Synta x BW = roicolor(A,low,high) BW = roicolor(A,v) Description roicolor selects a region of interest within an indexed or intensity image and returns a binary image. (You can use the returned image as a mask for masked filtering using roifilt2.) BW = roicolor(A,low,high) returns a region of interest selected as those pixels that lie within the colormap range [low high]: BW = (A >= low) & (A <= high) BW is a binary image with 0’s outside the region of interest and 1’s inside. BW = roicolor(A,v) returns a region of interest selected as those pixels in A that match the values in vector v. BW is a binary image with 1’s where the values of A match the values of v. Cla ss Support Ex a mple See Also The input image A can be of class uint8 or double. The output image BW is of class uint8. I = imread('rice.tif'); BW = roicolor(I,128,255); imshow(I); figure, imshow(BW) roifilt2, roipoly www.MatlabSite.com 11-181 roifill Purpose Synta x 11roifill Smoothly interpolate within an arbitrary image region J = roifill(I,c,r) J = roifill(I) J = roifill(I,BW) [J,BW] = roifill(...) J = roifill(x,y,I,xi,yi) [x,y,J,BW,xi,yi] = roifill(...) Description roifill fills in a specified polygon in an intensity image. It smoothly interpolates inward from the pixel values on the boundary of the polygon by solving Laplace’s equation. roifill can be used, for example, to “erase” small objects in an image. J = roifill(I,c,r) fills in the polygon specified by c and r, which are equal-length vectors containing the row-column coordinates of the pixels on vertices of the polygon. The k-th vertex is the pixel (r(k),c(k)). J = roifill(I) displays the image I on the screen and lets you specify the polygon using the mouse. If you omit I, roifill operates on the image in the current axes. Use normal button clicks to add vertices to the polygon. Pressing Backspace or Delete removes the previously selected vertex. A shift-click, right-click, or double-click adds a final vertex to the selection and then starts the fill; pressing Return finishes the selection without adding a vertex. J = roifill(I,BW) uses BW (a binary image the same size as I) as a mask. roifill fills in the regions in I corresponding to the nonzero pixels in BW. If there are multiple regions, roifill performs the interpolation on each region independently. [J,BW] = roifill(...) returns the binary mask used to determine which pixels in I get filled. BW is a binary image the same size as I with 1’s for pixels corresponding to the interpolated region of I and 0’s elsewhere. J = roifill(x,y,I,xi,yi) uses the vectors x and y to establish a nondefault spatial coordinate system. xi and yi are equal-length vectors that specify polygon vertices as locations in this coordinate system. 11-182 www.MatlabSite.com roifill [x,y,J,BW,xi,yi] = roifill(...) returns the XData and YData in x and y; the output image in J; the mask image in BW; and the polygon coordinates in xi and yi. xi and yi are empty if the roifill(I,BW) form is used. If roifill is called with no output arguments, the resulting image is displayed in a new figure. Cla ss Support Ex a mple See Also The input image I and the binary mask BW can be of class double or uint8. The output image J is of the same class as I. All other inputs and outputs are of class double. I = imread('eight.tif'); c = [222 272 300 270 221 194]; r = [21 21 75 121 121 75]; J = roifill(I,c,r); imshow(I) figure, imshow(J) roifilt2, roipoly www.MatlabSite.com 11-183 roifilt2 Purpose 11roifilt2 Filter a region of interest Synta x J = roifilt2(h,I,BW) J = roifilt2(I,BW,fun) J = roifilt2(I,BW,fun,P1,P2,...) Description J = roifilt2(h,I,BW) filters the data in I with the two-dimensional linear filter h. BW is a binary image the same size as I that is used as a mask for filtering. roifilt2 returns an image that consists of filtered values for pixels in locations where BW contains 1’s, and unfiltered values for pixels in locations where BW contains 0’s. For this syntax, roifilt2 calls filter2 to implement the filter. J = roifilt2(I,BW,fun) processes the data in I using the function fun. The result J contains computed values for pixels in locations where BW contains 1’s, and the actual values in I for pixels in locations where BW contains 0’s. fun can be a string containing the name of a function, a string containing an expression, or an inline function object. fun should take a matrix as a single argument and return a matrix of the same size: y = fun(x) J = roifilt2(I,BW,fun,P1,P2,...) passes the additional parameters P1,P2,..., to fun. Cla ss Support 11-184 For the syntax that includes a filter h, the input imageI can be of class double or uint8, and the output array J is of class double. For the syntax that includes a function, I can be of any class supported by fun, and the class of J depends on the class of the output from fun. www.MatlabSite.com roifilt2 Ex a mple This example continues the roipoly example. h = fspecial('unsharp'); J = roifilt2(h,I,BW); imshow(J) See Also filter2, roipoly www.MatlabSite.com 11-185 roipoly Purpose Synta x 11roipoly Select a polygonal region of interest BW = roipoly(I,c,r) BW = roipoly(I) BW = roipoly(x,y,I,xi,yi) [BW,xi,yi] = roipoly(...) [x,y,BW,xi,yi] = roipoly(...) Description Use roipoly to select a polygonal region of interest within an image. roipoly returns a binary image that you can use as a mask for masked filtering. BW = roipoly(I,c,r) returns the region of interest selected by the polygon described by vectors c and r. BW is a binary image the same size as I with 0’s outside the region of interest and 1’s inside. BW = roipoly(I) displays the image I on the screen and lets you specify the polygon using the mouse. If you omit I, roipoly operates on the image in the current axes. Use normal button clicks to add vertices to the polygon. Pressing Backspace or Delete removes the previously selected vertex. A shift-click, right-click, or double-click adds a final vertex to the selection and then starts the fill; pressing Return finishes the selection without adding a vertex. BW = roipoly(x,y,I,xi,yi) uses the vectors x and y to establish a nondefault spatial coordinate system. xi and yi are equal-length vectors that specify polygon vertices as locations in this coordinate system. [BW,xi,yi] = roipoly(...) returns the polygon coordinates in xi and yi. Note that roipoly always produces a closed polygon. If the points specified describe a closed polygon (i.e., if the last pair of coordinates is identical to the first pair), the length of xi and yi is equal to the number of points specified. If the points specified do not describe a closed polygon, roipoly adds a final point having the same coordinates as the first point. (In this case the length of xi and yi is one greater than the number of points specified.) [x,y,BW,xi,yi] = roipoly(...) returns the XData and YData in x and y; the mask image in BW; and the polygon coordinates in xi and yi. If roipoly is called with no output arguments, the resulting image is displayed in a new figure. 11-186 www.MatlabSite.com roipoly Cla ss Support The input image I can be of class double or uint8. The output image BW is of class uint8. All other inputs and outputs are of class double. Rema rk s For any of the roipoly syntaxes, you can replace the input image I with two arguments, m and n, that specify the row and column dimensions of an arbitrary image. For example, these commands create a 100-by-200 binary mask: c = [112 112 79 79]; r = [37 66 66 37]; BW = roipoly(100,200,c,r); If you specify m and n with an interactive form of roipoly, an m-by-n black image is displayed, and you use the mouse to specify a polygon within this image. Ex a mple See Also I = imread('eight.tif'); c = [222 272 300 270 221 194]; r = [21 21 75 121 121 75]; BW = roipoly(I,c,r); imshow(I) figure, imshow(BW) roifilt2, roicolor, roifill www.MatlabSite.com 11-187 std2 Purpose 11std2 Compute the standard deviation of the elements of a matrix Synta x b = std2(A) Description b = std2(A) computes the standard deviation of the values in A. Cla ss Support A is an array of class uint8 or double. b is a scalar of class double. Algorithm std2 computes the standard deviation of the array A using std(A(:)). See Also corr2, mean2 std, mean in the online MATLAB Function Reference 11-188 www.MatlabSite.com subim a ge Purpose 11subimage Display multiple images in the same figure Synta x subimage(X,map) subimage(I) subimage(BW) subimage(RGB) subimage(x,y,...) h = subimage(...) Description You can use subimage in conjunction with subplot to create figures with multiple images, even if the images have different colormaps. subimage works by converting images to truecolor for display purposes, thus avoiding colormap conflicts. subimage(X,map) displays the indexed image X with colormap map in the current axes. subimage(I) displays the intensity image I in the current axes. subimage(BW) displays the binary image BW in the current axes. subimage(RGB) displays the truecolor image RGB in the current axes. subimage(x,y...) displays an image using a nondefault spatial coordinate system. h = subimage(...) returns a handle to an image object. Cla ss Support The input image can be of class uint8 or double. www.MatlabSite.com 11-189 subima ge Ex a mple See Also load trees [X2,map2] = imread('forest.tif'); subplot(1,2,1), subimage(X,map) subplot(1,2,2), subimage(X2,map2) imshow subplot in the online MATLAB Function Reference 11-190 www.MatlabSite.com truesize Purpose 11truesize Adjust display size of an image Synta x truesize(fig,[mrows mcols]) truesize(fig) Description truesize(fig,[mrows ncols]) adjusts the display size of an image. fig is a figure containing a single image or a single image with a colorbar. [mrows ncols] is a 1-by-2 vector that specifies the requested screen area in pixels that the image should occupy. truesize(fig) uses the image height and width for [mrows ncols]. This results in the display having one screen pixel for each image pixel. If you omit the figure argument, truesize works on the current figure. Rema rk s If the'TruesizeWarning' toolbox preference is 'on', truesize displays a warning if the image is too large to fit on the screen. (The entire image is still displayed, but at less than true size.) If 'TruesizeWarning' is 'off', truesize does not display the warning. Note that this preference applies even when you call truesize indirectly, such as through imshow. See Also imshow, iptsetpref, iptgetpref www.MatlabSite.com 11-191 uint8 Purpose 11uint8 Convert data to unsigned 8-bit integers Synta x B = uint8(A) Description B = uint8(A) creates the unsigned 8-bit integer array B from the array A. If A is a uint8 array, B is identical to A. The elements of a uint8 array can range from 0 to 255. For any value in A outside this range, the resulting value in B is not defined (and may vary from platform to platform). The fractional part of each value in A is discarded on conversion. This means, for example, that uint8(102.99) is 102, not 103. Therefore, it is often a good idea to round off the values in A before converting to uint8. For example: B = uint8(round(A)) MATLAB supports these operations on uint8 arrays: • Displaying data values • Indexing into arrays using standard MATLAB subscripting • Reshaping, reordering, and concatenating arrays, using functions such as reshape, cat, and permute • Saving to and loading from MAT-files • The all and any functions • Logical operators and indexing • Relational operators MATLAB also supports the find function for uint8 arrays, but the returned array is of class double. Most of the functions in the Image Processing Toolbox accept uint8 input. See the individual Reference entries for information about uint8 support. Rema rk s uint8 is a MATLAB built-in function. See Also double, im2double, im2uint8 11-192 www.MatlabSite.com w a rp Purpose 11warp Display an image as a texture-mapped surface Synta x warp(X,map) warp(I,n) warp(BW) warp(RGB) warp(z,...) warp(x,y,z,...) h = warp(...) Description warp(X,map) displays the indexed image X with colormap map as a texture map on a simple rectangular surface. warp(I,n) displays the intensity image I with gray scale colormap of length n as a texture map on a simple rectangular surface. warp(BW) displays the binary image BW as a texture map on a simple rectangular surface. warp(RGB) displays the RGB image in the array RGB as a texture map on a simple rectangular surface. warp(z,...) displays the image on the surface z. warp(x,y,z...) displays the image on the surface (x,y,z). h = warp(...) returns a handle to a texture mapped surface. Cla ss Support The input image can be of class uint8 or double. Rema rk s Texture-mapped surfaces render more slowly than images. www.MatlabSite.com 11-193 w a rp Ex a mple This example texture maps an image of a test pattern onto a cylinder: [x,y,z] = cylinder; I = imread('testpat1.tif'); warp(x,y,z,I); See Also imshow image, imagesc, surf in the online MATLAB Function Reference 11-194 www.MatlabSite.com w iener2 Purpose 11wiener2 Perform two-dimensional adaptive noise-removal filtering Synta x J = wiener2(I,[m n],noise) [J,noise] = wiener2(I,[m n]) Description wiener2 lowpass filters an intensity image that has been degraded by constant power additive noise. wiener2 uses a pixel-wise adaptive Wiener method based on statistics estimated from a local neighborhood of each pixel. J = wiener2(I,[m n],noise) filters the image I using pixel-wise adaptive Wiener filtering, using neighborhoods of size m-by-n to estimate the local image mean and standard deviation. If you omit the [m n] argument, m and n default to 3. The additive noise (Gaussian white noise) power is assumed to be noise. [J,noise] = wiener2(I,[m n]) also estimates the additive noise power before doing the filtering. wiener2 returns this estimate in noise. Cla ss Support The input image I can be of class uint8 or double. The output image J is of the same class as I. Ex a mple Degrade and then restore an intensity image using adaptive Wiener filtering. I = imread('saturn.tif'); J = imnoise(I,'gaussian',0,0.005); K = wiener2(J,[5 5]); imshow(J) figure, imshow(K) Algorithm wiener2 estimates the local mean and variance around each pixel www.MatlabSite.com 11-195 w iener2 1 µ = ----------NM ∑ n 1, n 2 ∈ η 1 2 σ = ----------NM ∑ a ( n 1, n 2 ) 2 n 1, n 2 ∈ η a ( n 1, n 2 ) – µ 2 where η is the N-by-M local neighborhood of each pixel in the image A. wiener2 then creates a pixel-wise Wiener filter using these estimates 2 2 σ –ν - ( a ( n 1, n 2 ) – µ ) b ( n 1, n 2 ) = µ + ----------------2 σ where ν2 is the noise variance. If the noise variance is not given, wiener2 uses the average of all the local estimated variances. See Also filter2, medfilt2 Reference Lim, Jae S. Two-Dimensional Signal and Image Processing. Englewood Cliffs, NJ: Prentice Hall, 1990. pp. 536-540. 11-196 www.MatlabSite.com ycbcr2 rgb Purpose 11ycbcr2rgb Convert YCbCr values to RGB color space Synta x rgbmap = ycbcr2rgb(ycbcrmap) RGB = ycbcr2rgb(YCBCR) Description rgbmap = ycbcr2rgb(ycbcrmap) converts the YCbCr values in the colormap ycbcrmap to the RGB color space. If ycbcrmap is m-by-3 and contains the YCbCr luminance (Y) and chrominance (Cb and Cr) color components as its columns, then rgbmap is returned as an m-by-3 matrix that contains the red, green, and blue values equivalent to those colors. RGB = ycbcr2rgb(YCBCR) converts the YCbCr image YCBCR to the equivalent truecolor image RGB. Cla ss Support If the input is a YCbCr image, it can be of class uint8 or double; the output image is of the same class as the input image. If the input is a colormap, the input and output colormaps are both of class double. See Also ntsc2rgb, rgb2ntsc, rgb2ycbcr www.MatlabSite.com 11-197 zoom Purpose 11zoom Zoom in and out of an image Synta x zoom on zoom off zoom out zoom reset zoom zoom xon zoom yon zoom(factor) zoom(fig,option) Description zoom on turns on interactive zooming for the current figure. When zooming is enabled, clicking the mouse on a point within an axes changes the axes limits by a factor of 2, to either zoom in on or out from the point: • For a single-button mouse, zoom in by clicking the mouse button and zoom out by shift-clicking. • For a two- or three-button mouse, zoom in by clicking the left mouse button and zoom out by clicking the right mouse button. Clicking and dragging over an axes when interactive zooming is enabled draws a rubber-band box. When the mouse button is released, the axes zoom in to the region enclosed by the rubber-band box. Double-clicking within an axes returns the axes to its initial zoom setting. zoom out returns the plot to its initial zoom setting. zoom reset remembers the current zoom setting as the initial zoom setting. Later calls to zoom out, or double-clicks when interactive zoom mode is enabled, return to this zoom level. zoom toggles the interactive zoom status. zoom xon and zoom yon sets zoom on for the x- and y-axis, respectively. zoom(factor) zooms in by the specified factor, without affecting the interactive zoom mode. By default, factor is 2. A factor between 0 and 1 specifies zooming out by 1/factor. 11-198 www.MatlabSite.com zoom zoom(fig,option) applies the zoom command to the figure specified by fig. option is a string containing any of the above arguments. If you do not specify a figure, zoom works on the current figure. See Also imcrop www.MatlabSite.com 11-199 zoom 11-200 www.MatlabSite.com Index A adaptive filtering 7-23, 11-195 aliasing 3-5 analyzing images contour plots 7-7 edge detection 7-10, 11-52 histograms 7-8, 11-117 intensity profiles 7-4, 11-123 pixel values 7-3, 11-121 quadtree decomposition 7-11, 11-166 summary statistics 7-9 anti-aliasing 3-5, 11-129 applylut 8-20, 11-10 area of binary images 8-18, 11-16 arrays logical 1-7, 1-15 storing images 1-3 averaging filter 5-10, 11-70 B bestblk 11-12 bicubic interpolation 3-3 bilinear interpolation 3-3 binary image operations 8-2 connected-components labeling 8-15, 11-23 feature measurement 8-18 flood fill 8-13, 11-20 lookup-table operations 8-20, 11-10, 11-151 morphological operations 8-4, 11-25 neighborhoods 8-2, 11-10 object-based operations 8-10 padding borders 8-2 binary images 1-7, 11-147 4-connected neighborhoods 8-10 8-connected neighborhoods 8-10 converting from other types 11-96 displaying 2-6, 8-3 Euler number 8-19, 11-18 image area 8-18, 11-16 object selection 8-16, 11-30 perimeter determination 8-12, 11-29 processing 8-2 binary masks 9-3 blkproc 4-8, 11-13 block processing 4-2 block size 11-12 column processing 4-11 distinct blocks 4-8, 11-13 padding borders 4-5 sliding neighborhoods 4-4, 11-159 BMP files 1-11, 11-114, 11-126, 11-135 borders padding 4-5, 5-5, 8-2 brighten 11-15 bwarea 8-18, 11-16 bweuler 8-19, 11-18 bwfill 8-13, 11-20 bwlabel 8-15, 11-23 bwmorph 8-8, 11-25 bwperim 8-12, 11-29 bwselect 8-16, 11-30 C Canny edge detector 11-52 center pixel linear filtering 5-4 morphological operations 8-4 closure 8-7, 11-25 cmpermute 11-32 cmunique 11-33 col2im 11-34 www.MatlabSite.com I-1 Index colfilt 4-11, 11-35 color 10-2 approximation 10-6, 11-49, 11-102, 11-177 dithering 10-8, 11-49 quantization 10-6, 11-177 reducing number of colors 10-5 color depth 10-3 color spaces converting between 1-13, 10-9, 11-92, 11-160, 11-175, 11-178, 11-179, 11-197 HSV 10-10, 11-92 NTSC 10-9, 11-160, 11-178 RGB 10-9 YCbCr 10-10, 11-179, 11-197 colorbar 2-10, 11-37 colormap mapping 10-7 colormaps brightening 11-15 darkening 11-15 plotting RGB values 11-180 rearranging colors 11-32 reducing number of colors 10-7 removing duplicate entries 11-33 RGB components 1-5 column processing 4-11, 11-35 reshaping blocks into columns 11-97 reshaping columns into blocks 11-34 computational molecule 5-4 connected-components labeling 8-15, 11-23 contour plots 7-7, 11-103 conv2 5-3, 11-39 conversions between image types 1-12 convmtx2 11-41 convn 5-9, 11-42 convolution convolution matrix 11-41 Fourier transform 6-12 I-2 higher-dimensional 5-9, 11-42 separability 5-8 two-dimensional 5-3, 11-39, 11-62 convolution kernel 5-3 center pixel 5-4 coordinate systems pixel coordinates 1-16 spatial coordinates 1-17 corr2 7-9, 11-43 correlation 5-7, 11-62 Fourier transform 6-13 correlation coefficient 11-43 cropping an image 3-7, 11-105 D data types 8-bit integers 1-3, 1-13, 11-192 converting between 1-14, 11-50, 11-192 double-precision 1-3, 11-50 dct2 6-15, 11-44 dctmtx 6-16, 11-46 dilate 8-6, 11-47 dilation 8-4, 11-26, 11-47 discrete cosine transform 6-15, 11-44 image compression 6-17 inverse 11-93 transform matrix 6-16, 11-46 discrete Fourier transform 6-8 display techniques 2-2, 11-133 adding a colorbar 2-10, 11-37 displaying at true size 2-4, 11-191 multiple images 2-14, 11-189 preferences 2-3 texture mapping 2-17, 11-193 zooming 2-16, 11-198 www.MatlabSite.com Index distinct block operations 4-8 overlap 4-9, 11-13 zero padding 4-8 dither 1-12, 11-49 dithering 10-8, 11-49, 11-176 double 1-14, 11-50 E edge 7-10, 11-51 edge detection 7-10 methods 11-52 enhancing images intensity adjustment 7-14, 11-100 noise removal 7-20 erode 8-6, 11-56 erosion 8-4, 11-26, 11-56 Euclidean distance 7-4, 11-165 Euler number 8-19, 11-18 F fast Fourier transform 6-8 higher-dimensional 11-60 higher-dimensional inverse 11-95 two-dimensional 11-58 two-dimensional inverse 11-94 feature measurement 7-9, 11-108 binary images 8-18 fft 6-8 fft2 6-8, 11-58 fftn 6-8, 11-60 fftshift 11-61 file formats 1-11, 11-114, 11-126, 11-135 files displaying images in 2-8 reading image data from 11-126 reading image information from 11-114 writing image data to 11-135 filling a region 9-8 filter design 5-13 frequency sampling method 5-15, 11-67 frequency transformation method 5-14, 11-74 windowing method 5-16, 11-77, 11-81 filter2 5-7, 11-62 filters adaptive 7-23, 11-195 averaging 5-10, 11-70 binary masks 9-6 designing 5-13 Finite Impulse Response (FIR) 5-13 frequency response 5-19, 6-11 Gaussian 11-70 Infinite Impulse Response (FIR) 5-14 Laplacian 11-70 Laplacian of Gaussian 11-70 linear 5-3, 7-21, 11-62 median 7-21, 11-155 order-statistic 11-161 predefined types 5-9, 11-70 Prewitt 11-70 Sobel 5-11, 11-70 unsharp 11-70 FIR filters 5-13 flood-fill operation 8-13, 11-20 Fourier transform 6-3 computing frequency response 6-11 convolution 6-12 correlation 6-13 higher-dimensional 11-60 higher-dimensional inverse 11-95 rearranging output 11-61 two-dimensional 11-58 two-dimensional inverse 11-94 www.MatlabSite.com I-3 Index freqspace 5-18, 11-64 frequency response computing 5-19, 6-11, 11-65 desired response matrix 5-18, 11-64 frequency sampling method (filter design) 5-15, 11-67 frequency transformation method (filter design) 5-14, 11-74 freqz2 5-19, 6-11, 11-65 fsamp2 5-15, 11-67 fspecial 5-9, 11-70 ftrans2 5-14, 11-74 fwind1 5-17, 11-77 fwind2 5-17, 11-81 Hough transform detecting lines 6-22 HSV color space 10-10, 11-92, 11-175 hsv2rgb 10-10, 11-92 I idct2 11-93 ifft 6-8 ifft2 6-8, 11-94 ifftn 6-8, 11-95 IIR filters 5-14 im2bw 1-12, 11-96 im2col 11-97 im2double 1-14, 11-98 im2uint8 1-14, 11-99 G gamma correction 7-16 Gaussian filter 11-70 Gaussian noise 7-23 geometric functions 3-2 cropping 3-7, 11-105 interpolation 3-3 resizing 3-5, 11-129 rotation 3-6, 11-131 getimage 11-85 getting preference values 11-142 gray2ind 1-12, 11-87 grayscale morphological operations 11-161 grayslice 1-12, 11-88 H HDF files 1-11, 11-114, 11-126, 11-135 histeq 7-18, 11-89 histogram equalization 7-18, 11-89 histograms 7-8, 11-117 I-4 imadjust 7-14, 11-100 image area (binary images) 8-18, 11-16 image types binary 1-7, 8-2 converting between 1-11 indexed 1-5 intensity 1-6 multiframe images 1-9 RGB (truecolor) 1-8 images analyzing 7-3 color 10-2 color depth 10-3 converting to binary 11-96 data types 1-3, 11-50, 11-192 displaying 2-2, 11-133 displaying multiple images 2-14, 11-189 enhancing 7-14 file formats 1-11, 11-114, 11-126, 11-135 getting data from axes 11-85 mean value 11-154 www.MatlabSite.com Index reading data from files 11-126 reading information from files 11-114 reducing number of colors 10-5, 11-102 standard deviation 11-188 storing in MATLAB 1-3 writing to files 11-135 imapprox 10-7, 11-102 imcontour 7-7, 11-103 imcrop 3-7, 11-105 imfeature 7-9, 8-18, 11-108 imfinfo 1-11, 11-114 imhist 7-8, 11-117 immovie 2-13, 11-118 imnoise 7-21, 11-119 impixel 7-3, 11-121 improfile 7-4, 11-123 imread 1-11, 11-126 imresize 3-5, 11-129 imrotate 3-6, 11-131 imshow 2-3, 11-133 imwrite 1-11, 11-135 ind2gray 1-12, 11-139 ind2rgb 1-12, 11-140 indexed images 1-5, 11-149 converting from intensity 11-87 converting from RGB 11-176 converting to intensity 11-139 converting to RGB 11-140 displaying 2-5 reducing number of colors 10-5 intensity adjustment 7-14, 11-100 gamma correction 7-16 histogram equalization 7-18 intensity images 1-6, 11-148 converting from indexed 11-139 converting from matrices 11-153 converting from RGB 11-174 converting to indexed 11-87 displaying 2-5 number of gray levels displayed 2-6 intensity profiles 7-4, 11-123 interpolation 3-3 bicubic 3-3 bilinear 3-3 intensity profiles 7-4 nearest neighbor 3-3 iptgetpref 2-4, 11-141 iptsetpref 2-3, 11-142 iradon 6-25, 11-144 isbw 11-147 isgray 11-148 isind 11-149 isrgb 11-150 J JPEG files 1-11, 11-114, 11-126, 11-135 JPEG image compression 6-17 L Laplacian filter 11-70 Laplacian of Gaussian edge detector 11-52 Laplacian of Gaussian filter 11-70 linear filtering 4-5, 5-3, 11-62 averaging filter 5-10 center pixel 5-4 computational molecule 5-4 convolution 5-3 convolution kernel 5-3 correlation 5-7 filter design 5-13 FIR filters 5-13 IIR filters 5-14 www.MatlabSite.com I-5 Index noise removal 7-21 predefined filter types 5-9 Sobel filter 5-11 logical arrays 1-7, 1-15 lookup-table operations 8-20, 11-151 movies creating from images 2-13, 11-118 playing 2-14 multiframe images 1-9 displaying 2-11, 11-157 limitations 1-10 multilevel thresholding 11-88 M makelut 8-20, 11-151 masked filtering 9-6, 11-184 mat2gray 1-12, 11-153 matrices converting to intensity images 11-153 storing images 1-3 McClellan transform 11-74 mean2 7-9, 11-154 medfilt2 7-21, 11-155 median filtering 7-21, 11-155 minimum variance quantization 10-6, 11-177 Moiré patterns 3-5 montage 2-12, 11-157 morphological operations 8-4, 11-25 center pixel 8-4 closure 8-7, 11-25 diagonal fill 11-26 dilation 8-4, 11-26, 11-47 erosion 8-4, 11-26, 11-56 grayscale 11-161 opening 8-7, 11-26 predefined operations 8-8 removing spur pixels 11-27 shrinking objects 11-26 skeletonizing objects 11-26 structuring elements 8-4 thickening objects 11-27 thinning objects 11-27 I-6 N nearest neighbor interpolation 3-3 neighborhood operations 4-2 neighborhoods binary image operations 8-2, 8-10, 11-10 nlfilter 4-6, 11-159 noise removal 7-20 adding noise 11-119 Gaussian noise 7-23, 11-119 salt and pepper noise 7-21, 11-119 speckle noise 11-119 nonlinear filtering 4-5 NTSC color space 10-9, 11-160, 11-178 ntsc2rgb 10-9, 11-160 O object selection 8-16, 11-30 opening 8-7, 11-26 order-statistic filtering 11-161 ordfilt2 11-161 P padding borders binary image operations 8-2 block processing 4-5 linear filtering 5-5 www.MatlabSite.com Index PCX files 1-11, 11-114, 11-126, 11-135 perimeter determination 8-12, 11-29 phantom 6-25, 11-162 pixel coordinates 1-16 pixel values 7-3, 11-121, 11-165 pixels definition 1-3 pixval 7-4, 11-165 plotting colormap values 11-180 preferences 2-3 getting values 11-142 setting values 11-141 Prewitt edge detector 11-52 Prewitt filter 11-70 Q qtdecomp 7-11, 11-166 qtgetblk 11-169 qtsetblk 11-171 quadtree decomposition 7-11, 11-166 getting block values 11-169 setting block values 11-171 quantization minimum variance quantization 10-6, 11-177 uniform quantization 10-6, 11-177 R radon 6-19, 11-172 Radon transform 6-19, 11-172 detecting lines 6-22 inverse 6-25, 11-144 region of interest binary masks 9-3 filling 9-8, 11-182 filtering 9-6, 11-184 selecting 9-3, 9-5, 11-181, 11-186 resizing images 3-5, 11-129 anti-aliasing 3-5 RGB images 1-8, 11-150 converting from indexed 11-140 converting to indexed 11-176 converting to intensity 11-174 displaying 2-7 reducing number of colors 10-5 rgb2gray 1-12, 11-174 rgb2hsv 10-10, 11-175 rgb2ind 1-12, 10-5, 11-176 rgb2ntsc 10-9, 11-178 rgb2ycbcr 10-10 rgbplot 11-180 Roberts edge detector 11-52 roicolor 9-5, 11-181 roifill 9-8, 11-182 roifilt2 9-6, 11-184 roipoly 9-3, 11-186 rotating an image 3-6, 11-131 S salt and pepper noise 7-21 separability in convolution 5-8 setting preference values 11-141 sliding neighborhood operations 4-4, 11-159 Sobel edge detector 11-52 Sobel filter 5-11, 11-70 spatial coordinates 1-17 std2 7-9, 11-188 structuring elements 8-4 center pixel 8-4 subimage 2-16, 11-189 subplot 2-15 www.MatlabSite.com I-7 Index T Y template matching 6-13 texture mapping 2-17, 11-193 thresholding 11-88, 11-96 TIFF files 1-11, 11-114, 11-126, 11-135 transforms 6-2 discrete cosine 6-15, 11-44 Fourier 6-3, 11-58, 11-60, 11-61 inverse discrete cosine 11-93 inverse Fourier 11-94, 11-95 inverse Radon 6-25, 11-144 Radon 6-19, 11-172 truecolor images 1-8 truesize 2-4, 11-191 YCbCr color space 10-10, 11-179, 11-197 ycbcr2rgb 10-10 Z zero-cross edge detector 11-52 zoom 2-16, 11-198 zooming in 2-16, 11-198 U uint8 11-192 uint8 arrays storing images 1-3 supported operations 1-13, 11-192 uniform quantization 10-6, 11-177 unsharp filter 11-70 W warp 2-17, 11-193 wiener2 7-23, 11-195 windowing method (filter design) 5-16, 11-77, 11-81 X XWD files 1-11, 11-114, 11-126, 11-135 I-8 www.MatlabSite.com