Sketch-to-Sketch Match With Dice Similarity Measures

International Journal of Application or Innovation in Engineering & Management (IJAIEM)
Web Site: www.ijaiem.org Email: editor@ijaiem.org, editorijaiem@gmail.com Volume 3, Issue 2, February 2014 ISSN 2319 - 4847
Sketch-to-Sketch Match with Dice Similarity Measures

Mrs. Prajwalita Dongare1, Prof.P.D. Lambhate2
1
Computer Department, J.S.P.M Handewadi, Pune, Maharashtra 411028, India
HOD Information Technology, J.S.P.M Handewadi, Pune, Maharashtra 411028, India
Abstract
The most important thought of this project is to present competent, immediate, and accurate CBIR using simple and strong man machine interface by means of sketches. In other terms the aim is to locate content-based image features that are efficient enough enough to allow us to do sketch based image retrieval where a user-drawn sketch is applied as the starting point to search in an image dataset for conceptually similar images. This is a non trivial task since users usually do not have a good example at hand when they begin to search for images. Indexing ability of the proposed methods ensures their appropriateness in large-scale image image datasets. Our system improves the retrieval performance and better matches.
KeywordsSobel edge detector, EHD (edge histogram descriptor), haar wavelet, Indexing, Point Map algorithm, Sketch.
1. INTRODUCTION
In content based image queries, example images, unclear drawings of the desired colors; simple outline sketches; are recommended and analyze. We suppose that draw sketches are usually easier and quicker to generate than a complete color description of the picture. And they can be generated for subjective preferred images, while model images may or may not be at give when searching. In other terms, sketch-based image retrieval (SBIR) is a significant means of querying large image. Sketching is a universal form of communication. It seems that it is much easier for humans to sketch the outlines. Sketches do not contain any color hence it is binary. The Sketch-based image retrieval (SBIR) was introduced in QBIC and Visual SEEK systems. The Sketch Based Image Retrieval (SBIR) covenant through mapping of a user sketch to a similar image in the image dataset. The basic problem is SBIR is edge detection which is solved by using sobel edge detection algorithm. The Haar wavelet transform, in this application is one way of compressing images so they take fewer gaps when stored and transmitted. To improve the retrieval and matching performance, Edge Histogram Descriptor depicts edge allocation with a histogram based on neighboring edge allocation in an image.[1]-[2] Our system, reduce the physical size of the files without mortifying the superiority of the image to an undesirable level using image compression techniques. This leads to compressed version of an image, lessening in file size allows extra images to be stored in a certain amount of disk or memory space.
2. RELATED WORK
Content-based image retrieval (CBIR) supports both known image search and novel image search. However, in both cases CBIR requires a query image to start with that is similar to the final result. Without such a query image, it is difficult to achieve good retrieval quality. Query-by-Sketch addresses this problem and uses user drawn sketches as query images. Rajendran and Chang propose to extract edge signatures for both images and sketch query, and compare their curvature and direction histograms for shape similarity[2]. This method utilizes multiple scales to account for various levels of detail in sketch query. By contrast, the majority of early sketched retrieval work focused on the fitting of sketched contours directly to the image. Bimbo and Pala present a technique which based on elastic matching of sketched templates over edge maps in the image to evaluate similarity. The degree of matching achieved and the elastic deformation energy spent by the sketch to achieve such a match are used to derive a measure of similarity between the sketch and the images in the database and to rank images to be displayed. However the expense of the optimization step to fit the model limits the scalability of the approach, making it inappropriate for use in interactive SBIR systems. A. Sobel edge detector: Sobel operator is an operator used in image processing, mainly for edge detection algorithm. Actually, it is a distinct discrimination operator, computing a rough calculation of the gradient of the image strength function. It is also two
Volume 3, Issue 2, February 2014
Page 60

dimensional map of gradient at each point and can be progression and inspection as if it itself an image, with the area of high gradient or the likely edges visible as white lines. In the proposed method, Sobel mask is used to detect the shapes outer edges. The outer edge of each shape is needed to compute the boundary of each shape. The boundary is obtained by counting the total white pixels in the edge of a shape. At each image point, the gradient vector of the Sobel mask points add to in the direction of biggest possible intensity. The convolution between the Sobel operators with input image will produce edge, i.e. pixel values equal to one, where same value region will produce zeros and otherwise will produce ones.[4] B. Haar Wavelet Transform: Wavelets are mathematical tools for hierarchically decaying functions. In recent years Wavelet Transform has been proved to be a very helpful tool for image processing. The Haar Transform (HT) is one of the simplest and basic transformations from the space domain to a local frequency domain. A HT decomposes each signal into two components, one is called average and the other is known as difference. To understand averaging and differencing.[5] The formula for average is, av1 =(av1,av2,av3.avn/2) at one level of signal legth n i.e fq=(fq1,fq2,.fqn) fq is frequency. i.e avn= dfn= ,n=(1,2,3,.n/2).--------eq(1) ,n=(1,2,3,.n/2).--------eq(2). and the first detail sub-signal.df1=(df1,df2 ,df3..dfn/2). at the same level is given as
The haar wavelet transform produces four areas A, H, V and D respectively. A (approximation area) includes information about the global properties of analyzed image. Removal of spectral coefficients from this area leads to the biggest distortion in original image. H (horizontal area) includes information about the vertical lines hidden in image. Removal of spectral coefficients from this area excludes horizontal details from original image. V (vertical area) contains information about the horizontal lines hidden in image. Removal of spectral coefficients from this area eliminates vertical details from original image. D (diagonal area) embraces information about the diagonal details hidden in image. C. Edge Histogram Descriptor: Edge Histogram Descriptor describes edge distribution with a histogram based on local edge distribution in an image, used to measure similarity in images. The exact method of computation for the Edge Histogram descriptor can be found in and. The general idea is that the image is divided into 44 sub-images, and the local edge histograms are computed for each of the sub-images. There are 5 possible edge orientations that are considered: vertical, horizontal, 45 diagonal, 135 diagonal and isotropic (no orientation detected).For each sub-image and for each image type an edge intensity bin is computed, amounting to a total of 16 images 5edges=80bins.Each sub-image is further divided into sub-blocks, which are down-sampled into a 22 pixel image by intensity averaging, and the edge-detector operators are applied using the 5 filters in the image.[3]
3. SYSTEM OVERVIEW
In general, an image retrieval system generally provides a user interface for communicating with the user. It collects the necessary information, including the query image, from the user and displays the retrieval results. Our system work, in four steps i) Querying: The user give a sketch image as the query for the system. ii) Feature Extraction: The system extract the low level feature from the sketch image. The store images feature extraction is done using the EHD and stores into the array. iii) Similarity Measure: Comparison of drawn sketch and the stored images are through on feature extracted.
Figure 1 System Overview
Page 61

Web Site: www.ijaiem.org Email: editor@ijaiem.org, editorijaiem@gmail.com Volume 3, Issue 2, February 2014 ISSN 2319 - 4847 4. POINT MAP ALGORITHM
i. Input the image and convert to grayscale ii. Detect the edge of the image using canny detector iii. Store the valuable edge points in the buffer. iv. Used Haar wavelet transform for the sketched image to extract the contrast information. v. Transmit the valuable edge points and encoded information in the ideal noiseless channel. vi. Perform decoding, inverse transform at the restoration side. vii. Collect the edge in sequence from the buffer. viii. Trace the position of the edges. ix. Map the pixel positions of the edges and the transformed coefficients. Similarity Measures: The simplest similarity indices for binary images are based on shortest measurement of the degree of spatial overlap between these images. Similarity searching retrieves objects that are similar to a query, sorted in order of decreasing similarity. The Dice similarity coefficient is commonly used (e.g., in medical imaging studies) to quantify the degree of overlap between objects in two binary images. Assume that two objects or sketch, i and j are correspond to the binary feature vector form. Then similarity between the sketch image and images stored in the dataset are given by following formula: Dice is defined as Dice=(2a)/(2a+b+c)----------------------------------(3) Where, a is the number of attribute ,where the values of i and j are both 1,significance optimistic matches. b is the number of attributes where the value of i and j is (0,1), significance i dearth Mismatch. c is the number of attributes where the value of i and j is (1,0), significance j absence mismatch. The similarity in vector space representation is determined by using associative coefficients depends on the internal merchandise of the document vectors, where feature have common characteristics indicates similarity. The internal merchandise is usually standardized, since, in practice, not all features are equally applicable when assessing similarity. Naturally, a feature is more significant to a document if it appears more frequently in it than in the rest of documents. The binary similarity and distance measures play a vital role in sketch analysis problems such as classification, clustering, etc.[7]
5. EXPERIMENTAL SETUP AND RESULTS

Our Experiment is carried out on the image compression using haar wavelet and comparison with the EHD system. The compression technique is used to find the interested edge point required to our key value algorithm , after finding the key values of the its check with the stored images edges and find the similarity between them and display the result. Indexing is done on files which stored the related images and the features are stored in the featured array. Our system performance and retrieval of image is fast as compared to the existing system. The correctness of the correspondence between the query and retrieved image is taken as performance measure.
Figure 2 Indexing is the frist step.
Page 62

Figure 3 Edge detected using canny edge
Figure 4 Histrogram plotting
Page 63

Figure 5 Results obtained on basic shape with distance match
6. SEARCHING AND INDEXING ANALYSIS :

We had perfrom some test on the collection of different images set,time taken for searching and indexing the result on the desired qeury.The graph below shows the behaviour of the search reults from dataset and indexing results.
Figure 6 Retrieval time of Indexing and searching in number of images
Page 64

It shows that as number of images increases it takes time to indexing while in case of searching time is decreases because of the indexing is done before .
7. CONCLUSION AND FUTURE WORK

Retrieving images from a large collection of image dataset is still the challenge in the Content Based Image Retrieval. In Sketch Based Image Retrieval, a user sketch is mapped to a similar image in the image dataset. After feature extraction from the sketch image using our algorithm, and mapping it into the original image, using edge histogram descriptor, we found that the performance of the system is improved and the accuracy of the similar images is better. Currently our approach is to provide user to sketch freely and finding the similar match. Our system is not detecting any object .The future work will be on object detection from the images and overlapped images.
References
[1] Eitz, Mathias, et al. "Sketch-based shape retrieval." ACM Transactions on Graphics (TOG) 31.4 (2012): 31. [2] Kristian Hildebrand, Tamy Boubekeur, et al. "An evaluation of descriptors for large-scale image retrieval from sketched feature lines." Computers & Graphics 34.5 (2010): 482-498. [3] Chih-Chin Lai, Member, IEEE, and Ying-Chuan Chen A User-Oriented Image Retrieval System Based on Interactive Genetic Algorithm ,IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 60, NO. 10, OCTOBER 2011 [4] Pinaki Pratim Acharjya , Ritaban Das & Dibyendu Ghoshal Study and Comparison of Different Edge Detectors for Image Segmentation, Global Journal of Computer Science and Technology Graphics & Vision Volume 12 Issue 13 Version 1.0 Year 2012 Type: Double Blind Peer Reviewed International Research Journal Publisher: Global Journals Inc. (USA) Online ISSN: 0975-4172 & Print ISSN: 0975-4350 [5] Lai, Yung-Kai, and C-C. Jay Kuo. "A Haar wavelet approach to compressed image quality measurement." Journal of Visual Communication and Image Representation 11.1 (2000): 17-40. [6] A.Benjamin Joseph & Dr.R.Baskaran. Edge Preserved Image Compression Technique Using Wavelet and Edge Based Segmentation, International Journal of Computer Science & Informatics, Volume-I, Issue-II, 2011 [7] Seung-Seok Choi; Sung-Hyuk Cha; Tappert, A Survey of Binary Similarity and Distance MeasuresCharles Journal of Systemics, Cybernetics & Informatics . 2010.
Page 65

Sketch-to-Sketch Match With Dice Similarity Measures

Uploaded by

Copyright:

Available Formats

Sketch-to-Sketch Match With Dice Similarity Measures

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Sketch-to-Sketch Match With Dice Similarity Measures

Uploaded by

Copyright:

Available Formats

International Journal of Application or Innovation in Engineering & Management (IJAIEM)