Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.

5(22)

Content Based Image Retrieval

Srinivasa Kumar Devireddy
Dept. of Computer Science & Engineering, Nalanda Institute of Engineering & Technology,
Siddharth Nagar, Kantepudi(V), Sattenapalli(M), Guntur (Dt.), A.P.,India,
E-Mail: srinivaskumar_d@yahoo.com

Abstract
The importance of an effective technique in searching and retrieving images from
the huge collection cannot be overemphasized. One approach for indexing and
retrieving image data is using manual text annotations. The annotations can then be
used to search images indirectly. But there are several problems with this approach.
First, it is very difficult to describe the contents of an image using only a few keywords.
Second, the manual annotation process is very subjective, ambiguous, and incomplete.
Those problems have created great demands for automatic and effective techniques for
content-based image retrieval (CBIR) systems. Most CBIR systems use low-level image
features such as color, texture, shape, edge, etc., for image indexing and retrieval. Its
because the low-level features can be computed automatically. Content Based Image
Retrieval (CBIR) has emerged during the last several years as a powerful tool to
efficiently retrieve images visually similar to a query image. The main idea is to
represent each image as a feature vector and to measure the similarity between images
with distance between their corresponding feature vectors according to some metric.
Finding the correct features to represent images with, as well as the similarity metric
that groups visually similar images together, are important steps in the construction of
any CBIR system.

1 INTRODUCTION
The field of image retrieval has been an active research area for several decades and has been
paid more and more attention in recent years as a result of the dramatic and fast increase in the
volume of digital images. The development of Internet not only cause an explosively growing
volume of digital images, but also give people more ways to get those images.
There were two approaches to content-based image retrieval initially. The first one is based
on attribute representation proposed by database researchers where image contents are defined as a
set of attributes which are extracted manually and are maintained within the framework of
conventional database management systems. Queries are specified using these attributes. This
obviously involves high-level of image abstraction. The second approach which was presented by
image interpretation researchers depends on an integrated feature-extraction / object-recognition
subsystem to overcome the limitations of attribute-based retrieval. This system automates the
feature-extraction and object- recognition tasks that occur when an image is inserted into the
database. This automated approaches to object recognition are computationally expensive, difficult
and tend to be domain specific. There are two major categories of features. One is basic which is
concerned with extracting boundaries of the image and the other one is logical which defines the
image at various levels of details. Regardless of which approach is used, the retrieval in contentbased image retrieval is done by color, texture, sketch, shape, volume, spatial constraints, browsing,
objective attributes, subjective attributes, motion, text and domain concepts.
1.1 The growth of digital imaging: The use of images in human communication is hardly
new our cave-dwelling ancestors painted pictures on the walls of their caves, and the use of maps
and building plans to convey information almost certainly dates back to pre-Roman times. But the
10

Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.5(22)

twentieth century has witnessed unparalleled growth in the number, availability and importance of
images in all walks of life. Images now play a crucial role in fields as diverse as medicine,
journalism, advertising, design, education and entertainment. Technology, in the form of inventions
such as photography and television, has played a major role in facilitating the capture and
communication of image data. But the real engine of the imaging revolution has been the computer,
bringing with it a range of techniques for digital image capture, processing, storage and
transmission which would surely have started even pioneers like John Logie Baird. The
involvement of computers in imaging can be dated back to 1965, with Ivan Sutherlands Sketchpad
project, which demonstrated the feasibility of computerised creation, manipulation and storage of
images, though the high cost of hardware limited their use until the mid-1980s. Once computerised
imaging became affordable, it soon penetrated into areas traditionally depending heavily on images
for communication, such as engineering, architecture and medicine. Photograph libraries, art
galleries and museums, too, began to see the advantages of making their collections available in
electronic form. The creation of the World-Wide Web in the early 1990s, enabling users to access
data in a variety of media from anywhere on the planet, has provided a further massive stimulus to
the exploitation of digital images. The number of images available on the Web was recently
estimated to be between 10 and 30 million a figure which some observers consider to be a
significant underestimate.
1.2 CBIR: Content Based Image Retrieval is the retrieval of images based on visual features
such as colour and texture. Reasons for its development are that in many large image databases,
traditional methods of image indexing have proven to be insufficient, laborious, and extremely time
consuming. These old methods of image indexing, ranging from storing an image in the database
and associating it with a keyword or number, to associating it with a categorized description, have
become obsolete. This is not in CBIR. In CBIR, each image that is stored in the database has its
features extracted and compared to the features of the query image. It involves two steps:
Feature Extraction: The first step in the process is extracting image features to a
distinguishable extent.
Matching: The second step involves matching these features to yield a result that is visually
similar.
1.3 Existing System
Early techniques were not generally based on visual features but on the textual annotation of
images. In other words, images were first annotated with text and then searched using a text-based
approach from traditional database management systems. Text-based image retrieval uses
traditional database techniques to manage images. Through text descriptions, images can be
organized by topical or semantic hierarchies to facilitate easy navigation and browsing based on
standard Boolean queries. However, since automatically generating descriptive texts for a wide
spectrum of images is not feasible, most text-based image retrieval systems require manual
annotation of images. Obviously, annotating images manually is a cumbersome and expensive task
for large image databases, and is often subjective, context-sensitive and incomplete. As a result, it is
difficult for the traditional text-based methods to support a variety of task-dependent queries.
1.4 Proposed System
The solution proposed is to extract the primitive features of a query image and compare them
to those of database images. The image features under consideration were colour, texture and shape.
Thus, using matching and comparison algorithms, the colour, texture and shape features of one
image are compared and matched to the corresponding features of another image. This comparison
is performed using colour, texture and shape distance metrics. In the end, these metrics are
performed one after another, so as to retrieve database images that are similar to the query. The
similarity between features is to be calculated using Distance algorithm.
11

Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.5(22)

2 Applications of CBIR
A wide range of possible applications for CBIR technology has been identified. Potentially
fruitful areas include: Crime prevention, The military ,Fashion and interior design, Journalism and
advertising, Medical diagnosis , Geographical information and remote sensing systems , Web
searching.
Crime prevention: Law enforcement agencies typically maintain large archives of visual
evidence, including past suspects facial photographs (generally known as mug shots), fingerprints,
tyre treads and shoeprints. Whenever a serious crime is committed, they can compare evidence from
the scene of the crime for its similarity to records in their archives. Strictly speaking, this is an
example of identity rather than similarity matching, though since all such images vary naturally
over time, the distinction is of little practical significance. Of more relevance is the distinction
between systems designed for verifying the identity of a known individual (requiring matching
against only a single stored record), and those capable of searching an entire database to find the
closest matching records.
The military: Military applications of imaging technology are probably the best-developed,
though least publicized. Recognition of enemy aircraft from radar screens, identification of targets
from satellite photographs, and provision of guidance systems for cruise missiles are known
examples though these almost certainly represent only the tip of the iceberg. Many of the
surveillance techniques used in crime prevention could also be relevant to the military field.
Fashion and interior design: Similarities can also be observed in the design process in other
fields, including fashion and interior design. Here the designer has to work within externallyimposed constraints, such as choice of materials. The ability to search a collection of fabrics to find
a particular combination of colour or texture is increasingly being recognized as a useful aid to the
design process.
Journalism and advertising :Both newspapers and stock shot agencies maintain archives of
still photographs to illustrate articles or advertising copy. These archives can often be extremely
large (running into millions of images), and dauntingly expensive to maintain if detailed keyword
indexing is provided. Broadcasting corporations are faced with an even bigger problem, having to
deal with millions of hours of archive video footage, which are almost impossible to annotate
without some degree of automatic assistance.
Medical diagnosis :The increasing reliance of modern medicine on diagnostic techniques
such as radiology, histopathology, and computerised tomography has resulted in an explosion in the
number and importance of medical images now stored by most hospitals. While the prime
requirement for medical imaging systems is to be able to display images relating to a named patient,
there is increasing interest in the use of CBIR techniques to aid diagnosis by identifying similar past
cases.
Geographical information systems (GIS) and remote sensing : Although not strictly a
case of image retrieval, managers responsible for planning marketing and distribution in large
corporations need to be able to search by spatial attribute (e.g. to find the 10 retail outlets closest to
a given warehouse). And the military are not the only group interested in analysing satellite images.
Agriculturalists and physical geographers use such images extensively, both in research and for
more practical purposes, such as identifying areas where crops are diseased or lacking in nutrients
or alerting governments to farmers growing crops on land they have been paid to leave lying fallow.
Web searching: Cutting across many of the above application areas is the need for effective
location of both text and images on the Web, which has developed over the last five years into an
12

Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.5(22)

indispensable source of both information and entertainment. Text-based search engines have grown
rapidly in usage as the Web has expanded; the well-publicized difficulty of locating images on the
Web indicates that there is a clear need for image search tools of similar power. Paradoxically, there
is also a need for software to prevent access to images which are deemed pornographic.
3 Literature Survey
3.1 Image and Digital Imaging
It is said that one image is worth a thousand words. Visual information accounts for about
90% of the total information content that a person acquires from the environment through his
sensory systems. This reflects the fact that human being relies heavily on his highly developed
visual system compared with other sensory pathways. The external optical signal is perceived by
eyes, and then converted into neural signal; the corresponding neural subsystem specialized for
visual system is specially organized to detect subtle image features and perform high-level
processing, which is further processed to generate object entities and concepts. The anatomy of the
visual system explains from the structure aspect why visual information is so important to human
cognition. The cognitive functions that such a system must support include the capability to
distinguish among objects, their positions in space, motion, sizes, shapes, and surface texture. Some
of these primitives can be used as descriptors of image content in machine vision research.
3.1.1 Taxonomy of Images
There are two formats, in which visual information can be recorded and presented static
image, and motion picture, or video. Image is the major focus of research interest in digital image
processing and image understanding. Although a relatively recent development, computerized
digital image processing has attracted much attention and shed lights to a broad range of existing
and potential applications. This is directly caused by rapid accumulation of image data, a
consequence of exponential increases of digital storage capacity and computer processing power.
There are several major types of digital images depending on the elemental constituents that convey
the image content. Images can take the form of:
1. Printed text and manuscript. Some examples of the kind are micro-films of old text
documents, photograph of handwriting.
2. Line sketch, including diagrams, simple line graphs
3. Halftones. Images are represented by a grid of dots of variable sizes and shapes.
4. Continuous tone. Photographic images that use smooth and subtle tones.
5. Mixture of above.
Among all above, continuous tone or photographic images are most common in the digital
imaging practice and of major concern of content-based image retrieval. They used to be generated
by converting from other media using a scanner. The process is laboured intensive and costly. It
was estimated that the image capturing and the subsequent manual indexing may account for 90
percent of the total cost of building an image database. This was the situation a few years ago. Now,
digital cameras are becoming very popular and images are also being converted from analog
electronic format to digital format.
At the conceptual level, an image is a representation of its target object(s). According to the
Webster's 3rd New International Dictionary, an image is the optical counterpart of an object ... a
mental picture, a mental conception ... The definition reflects the fact that an image has its content,
which captures the optical or mental properties of an object; with its format varying across different
kinds of media. It is determined by how the optical properties are quantized, or the degree of mental
abstractions that is required. Here are some examples of image according to the above broader
definition:
1. An image can be an array of pixel values stored in uncompressed bitmap digital image
format. In this format, each value represents the color intensity at discrete points or pixels. A wellknown example of this is Microsoft's BMP format. Although BMP format allows for pixel packing
13

Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.5(22)

and run-length encoding to achieve certain level of compression, its uncompressed version is more
popular.
2. Popular Internet standard image formats see more extensive image transformation and
compression, such as GIF and JPEG. The GIF standard defines a color degeneration process, which
maps the colors in an image into no more than 256 new colors.
3. Specially defined signature file stores specifically extract images features in numeric or
Boolean format, indicating the presence (or non-presence) and strength of the features. This is a
very compact representation of image that is targeted for fast retrieval instead of display, archival,
etc. Only the essential image features are conserved in the signature file and they are algorithm
dependent. This allows for easy indexing and fast search for matching features of the query. An
example of this can be found in image coding method using vector quantization (VQ), in which
image blocks are coded according to a carefully chosen codebook. If the image blocks are similar to
each other or the images in a set bear significant similarity, higher compression ratio can usually be
achieved than the general- purpose compression algorithms such as GIF, JPEG.
4. Textual annotation can also be thought of as an instantiation of mental image, and
sometimes, the descriptors can be coded by a predefined convention, or a thesaurus. The fact that
two visually different images can convey the same concept and different concepts may present in
images that share many similar optimal properties brings about a gap between image retrieval by
content, and retrieval by concept.
3.2 Visual Features
Feature extraction plays an important role in content-based image retrieval to support for
efficient and fast retrieval of similar images from image databases. Significant features must first be
extracted from image data. Retrieving images by their content, as opposed to external features, has
become an important operation. A fundamental ingredient for content based image retrieval is the
technique used for comparing images. There are two general methods for image Comparison:
intensity based (color and texture) and geometry based (shape). So we will concentrate on these
features
3.2.1 Colour
One of the most important features that make possible the recognition of images by humans is
colour. Colour is a property that depends on the reflection of light to the eye and the processing of
that information in the brain. We use colour everyday to tell the difference between objects, places,
and the time of day. Usually colours are defined in three dimensional colour spaces. These could
either be RGB (Red, Green, and Blue), HSV (Hue, Saturation, and Value) or HSB (Hue, Saturation,
and Brightness). The last two are dependent on the human perception of hue, saturation, and
brightness. Most image formats such as JPEG, BMP, GIF, use the RGB colour space to store
information. The RGB colour space is defined as a unit cube with red, green, and blue axes. Thus, a
vector with three co-ordinates represents the colour in this space. When all three coordinates are set
to zero the colour perceived is black. When all three coordinates are set to 1 the colour perceived is
white. The other colour spaces operate in a similar fashion but with a different perception.
3.2.2 Transformation and Quantization
The color regions are perceptually distinguishable to some extent. The human eye cannot
detect small color different and may perceive these very similar colors as the same color. This leads
to the quantization of color, which means that some pre-specified colors will be present on the
image and each color is mapped to some of these pre-specified colors. One obvious consequence of
this is that each color space may require different levels of quantized colors, which is nothing but a
different quantization scheme. In below, the effect of color quantization is illustrated. Figure (a) is
the original image with RGB color space and (b) is the image produced after transformation into
HSV color space and quantization. A detailed explanation of color space transformations (from RGB
into HSV ) and quantization can be found in
14

Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.5(22)

Figure 1) Transformation, Quantization of tiger Image. (a) Original image. (b) Image
produced by applying RGB to HSV color transformation and quantization.
3.2.3 Methods of Representation
The main method of representing colour information of images in CBIR systems is through
colour histograms. A colour histogram is a type of bar graph, where each bar represents a particular
colour of the colour space being used. We can get a colour histogram of an image in the RGB or
HSV colour space. The bars in a colour histogram are referred to as bins and they represent the xaxis. The number of bins depends on the number of colours there are in an image. The y-axis
denotes the number of pixels there are in each bin. In other words how many pixels in an image are
of a particular colour.
3.2.4 Color Content Extraction
One of the widely used methods for querying and retrieval by color content is color
histograms. The color histograms are used to represent the color distribution in an image mainly,
the color histogram approach counts the number of occurrences of each unique color on a sample
image. Since an image is composed of pixels and each pixel has a color, the color histogram of an
image can be computed easily by visiting every pixel once. Smith and Chang proposed colorsets as
an opponent to color histograms. The colorsets are binary masks on color histograms and they store
the presence of colors as 1 without considering their amounts. For the absent colors, the colorsets
store 0 in the corresponding bins. The colorsets reduce the computational complexity of the distance
between two images. Besides, by employing colorsets region-based color queries are possible to
some extent. On the other hand, processing regions with more than two or three colors is quite
complex. Another image content storage and indexing mechanism is color correlograms. It
involves an easy-to-compute method and includes not only the spatial correlation of color regions
but also the global distribution of local spatial correlation of colors. In fact, a color correlogram is a
table each row of which is for a specific color pair of an image. The k-th entry in a row for color
pair (i; j) is the probability of finding a pixel of color j at a distance k from a pixel of color i. The
method resolves the drawbacks of the pure local and pure global color indexing methods since it
includes local spatial color information as well as the global distribution of color information.
3.2.5 Texture
Texture is that innate property of all surfaces that describes visual patterns, each having
properties of homogeneity. It contains important information about the structural arrangement of the
surface, such as; clouds, leaves, bricks, fabric, etc. It also describes the relationship of the surface to
the surrounding environment .In short, it is a feature that describes the distinctive physical
composition of a surface. Texture properties include: (i) Coarseness, ii)Contrast, iii)Directionality,
iv)Line-likeness, v)Regularity & v)Roughness.

15

Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.5(22)

(c) Rocks
(a) Clouds
(b) Bricks

Figure 2) Examples of Textures

Texture is one of the most important defining features of an image. It is characterized by the
spatial distribution of gray levels in a neighbourhood. In order to capture the spatial dependence of
gray-level values, which contribute to the perception of texture, a two-dimensional dependence
texture analysis matrix is taken into consideration. This two-dimensional matrix is obtained by
decoding the image file; jpeg, bmp, etc.
3.3 MPEG-7 Texture Descriptors
The MPEG-7 multimedia content description interface involves three texture descriptors for
representing texture regions in images, namely
Texture browsing descriptor to characterize perceptual directionality, regularity, and
coarseness of a texture,
Homogeneous texture descriptor (HTD) to quantitatively characterize homogeneous
texture regions for similarity retrieval using local spatial statistics of the texture obtained
by scale- and orientation-selective Gabor filtering, and
Edge histogram descriptor to characterize non-homogeneous texture regions.

3.4 Distance Measure Techniques

3.4.1 Histogram Intersection Method
In the Histogram Intersection technique, two normalized histograms are intersected as a
whole, as the name of the technique implies. The similarity between the histograms is a floating
point number between 0 and 1. Equivalence is designated with similarity value 1 and the similarity
between two histograms decreases when the similarity value approaches to 0. Both of the
histograms must be of the same size to have a valid similarity value. Let H1[1::n] and H2[1::n]
denote two histograms of size n, and SH1;H2 denote the similarity value between H1 and H2. Then,
SH1;H2 can be expressed by the distance between the histograms H1 and H2 as:

In the system, this technique is employed for similarity calculations as a result of texture
vector and color histogram comparisons between database images and query image.

16

Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.5(22)

4 General schema of Content Based image Retrieval
The block diagram consists of following main blocks - digitizer, feature extraction, image
database, feature database, and matching and multidimensional indexing. Function of each block is
as follows.

Fig 3) General Scheme of Content Based image Retrieval

Digitizer: To add new images in image database or query images which are acquired from
CCD Camera, X-ray imaging system, microdensitometers, image dissectors, vision cameras etc. are
needed to be digitized, so that computer can process those images.
Image Database: The Comparison between Query image and images from image database
can be done directly pixel by pixel which will give precise match but on the other hand, recognizing
objects entirely at query time will limit the retrieval speed of the system, due to the high expense of
such computing. Generally this crude method of comparison is not used, but image database, which
contains raw images, is required for visual display purpose.
Feature Extraction: To avoid above problem of pixel-by-pixel comparison next abstraction
level for representing images is the feature level. Every image is characterized by a set of features
such as Texture, Color, Shape and others. Extract these features at the time of injecting new image
in image database. Then summarize these features in a reduced set of k indexes and store it in
Image feature database. The query image is processed in the same way as images in the database.
Matching is carried out on the feature database.
Image matching and Multidimensional Indexing: Extracted features of query image are
compared with features, which are stored in image feature database. To achieve fast retrieval speed
and make the retrieval system truly scalable to large size image collections an effective
multidimensional indexing is indispensable part of the whole system. The system selects the N
images having the greatest overall similarities to the query image.
5 Theoretical Analysis
The basic idea behind content- based image retrieval is that, when building an image
database, or retrieving an image from the database, we first extract feature vectors from images (the
17

Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.5(22)

features can be color, shape, and texture), then store the vectors in another database for future use.
When given a query image, we similarly extract its feature vectors, and match these vectors with
those already in the database, if the distance between two images feature vectors is small enough;
we consider the corresponding image in the database match the query. The search is usually based
on similarity rather than on exact match, and the retrieval results are then ranked according to a
similarity index. Usually, a group of similar target images are presented to users.
Recent efforts in image retrieval applications have focused on indexing a few specific visual
dimensions of the images, such as color, texture, shape, motion and spatial information. However
without integrating these visual dimensions, the current content-based techniques have limited
capacity to satisfactorily retrieve images. Many difficult problems in image retrieval systems
remain to be investigated. The explosive proliferation of unconstrained digital imagery in the
form of images, graphics, and videos, due to the improved accessibility of computer technology and
the main-stream acceptance of the world-Wide Web as a variable medium for publishing,
advertising and communicating has created an immediate need for efficient and effective tools for
cataloguing, indexing, managing, compressing and searching for the unconstrained visual
information. A significant gap exists between the ability of computers to analyze images and videos
at the feature level (colors, textures, shapes) compared to the inability at the semantic-level (objects,
scenes, people, moods, artistic value). Closing this gap by improving technologies for image
understanding would certainly improve the image search and retrieval systems. However, the
solution of this large problem remains distant while much of the current technology consists of
solutions to small problems (for example, image segmentation, color constancy, shape from
texture, video shot detection, and so forth). However there are new opportunities to climb the
semantic ladder in todays divers media enjoinments. The final difficulty limiting progress in image
retrieval concerns system evaluation. Unless there are reliable and widely accepted ways of
measuring effectiveness of new technique, it will be impossible to judge whether they represent any
advancement on existing methods. This will inventible limit the progress. For an image retrieval
system to be successful for geographical images, we need to develop approaches robust in spatial
pattern characterization, rotation invariant, and taking into account of the scale effect. This project
can be thought of having two phases of operations. The first one consists of different types of image
storage techniques, the second one is having the retrieving the image from existing database.
What kinds of query are users likely to put to an image database? To answer this question in
depth requires a detailed knowledge of user needs why users seek images, what use they make of
them, and how they judge the utility of the images they retrieve. As we show in below, not enough
research has yet been reported to answer these questions with any certainty. Common sense
evidence suggests that still images are required for a variety of reasons, including:
illustration of text articles, conveying information or emotions difficult to describe in
words,
display of detailed data (such as radiology images) for analysis,
formal recording of design data (such as architectural plans) for later use.
Access to a desired image from a repository might thus involve a search for images depicting
specific types of object or scene, evoking a particular mood, or simply containing a specific texture
or pattern. Potentially, images have many types of attribute which could be used for retrieval,
including:
the presence of a particular combination of colour, texture or shape features (e.g. green
stars);
the presence or arrangement of specific types of object (e.g. chairs around a table);
the depiction of a particular type of event (e.g. a football match);
the presence of named individuals, locations, or events (e.g. the Queen greeting a
crowd);
subjective emotions one might associate with the image (e.g. happiness);
18

Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.5(22)

metadata such as who created the image, where and when.
Level 1 comprises retrieval by primitive features such as colour, texture, shape or the spatial
location of image elements. Examples of such queries might include find pictures with long thin
dark objects in the top left-hand corner, find images containing yellow stars arranged in a ring
or most commonly find me more pictures that look like this. This level of retrieval uses features
which are both objective, and directly derivable from the images themselves, without the need to
refer to any external knowledge base. Its use is largely limited to specialist applications such as
trademark registration, identification of drawings in a design archive, or colour matching of fashion
accessories.
Level 2 comprises retrieval by derived (sometimes known as logical) features, involving some
degree of logical inference about the identity of the objects depicted in the image. It can usefully be
divided further into:
1. retrieval of objects of a given type (e.g. find pictures of a double-decker bus);
2. retrieval of individual objects or persons (find a picture of the Eiffel tower).
To answer queries at this level, reference to some outside store of knowledge is normally
required particularly for the more specific queries at level 2(b). In the first example above, some
prior understanding is necessary to identify an object as a bus rather than a lorry; in the second
example, one needs the knowledge that a given individual structure has been given the name the
Eiffel tower. Search criteria at this level, particularly at level, are usually still reasonably objective.
Level 3 comprises retrieval by abstract attributes, involving a significant amount of high-level
reasoning about the meaning and purpose of the objects or scenes depicted. Again, this level of
retrieval can usefully be subdivided into:
1. retrieval of named events or types of activity (e.g. find pictures of Scottish folk
dancing);
2. retrieval of pictures with emotional or religious significance (find a picture depicting
suffering).
Success in answering queries at this level can require some sophistication on the part of the
searcher. Complex reasoning, and often subjective judgement, can be required to make the link
between image content and the abstract concepts it is required to illustrate. Queries at this level,
though perhaps less common than level 2, are often encountered in both newspaper and art libraries.

Fig 4) Architecture of CBIR

6 IMPLEMENTATION DETAILS
6.1 Color Histogram calculation
We define color histograms as a set of bins where each bin denotes the probability of pixels in
the image being of a particular color. A color histogram H for a given image is defined as a vector:
H = {H[0],H[1],H[2],..,H[i],H[n]}
19

Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.5(22)

Where i represents color in the color histogram[i] is the Number of pixels in color i in that
image and n is the number of bins in the color histogram.Typically, each pixel in an image will be
assigned to a bin of a color histogram of that image, so for the color histogram of an image, the
value of each bin is the number of pixels that has the same corresponding color. In order to compare
images of different sizes, color histograms should be normalized. The normalized color histogram
H is defined as:
H = {H[0],H[1],H[2],..,H[i],H[n]} Where H[i]=H[i]/p, p is the total number of
pixels in an image. An ideal color space quantization presumes that distinct colors should not be
located in the same sub-cube and similar colors should be assigned to the same sub-cube. Using few
colors will decrease the possibility that similar colors are assigned to different bins, but it increases
the possibility that distinct colors are assigned to the same bins, and that the information content of
the images will decrease by a greater degree as well. On the other hand, color histograms with a
large number of bins will contain more information about the content of images, thus decreasing the
possibility of distinct colors will be assigned to the same bins. However, they increase the
possibility that similar colors will be assigned to different bins, the storage space of metadata, and
the time for calculating the distance between color histograms. Therefore, there is a trade-off in
determining how many bins should be used in color histograms.
6.2. Converting RGB to HSV
The value is given by V =

R+G +B
.
3

Where the quantities R, G and B are the amounts of the red, green and blue Components
normalized to the range [0; 1]. The value is therefore just the average of the red, green and blue
components.
The saturation is given by S = 1

min( R , G , B )
3
=1
min( R , G , B ) .
I
R+G + B

Where the min(R; G;B) term is really just indicating the amount of white Present. If any of R,
G or B are zero, there is no white and we have a pure colour. The hue is given by
H = cos

1
[( R G ) + ( R B ) ]

( R G ) 2 + ( R B )(G B )

1
2

6.3. Edge Histogram calculation

Edges in images constitute an important feature to represent their content. Also, human eyes
are sensitive to edge features for image perception. One way of representing such an important edge
feature is to use a histogram. An edge histogram in the image space represents the frequency and
the directionality of the brightness changes in the image. It is a unique feature for images. To
represent this unique feature, in MPEG-7, there is a descriptor for edge distribution in the image.
This Edge Histogram Descriptor (EHD) proposed for MPEG-7 expresses only the local edge
distribution in the image. That is, since it is important to keep the size of the descriptor as compact
as possible for efficient storage of the metadata, the MPEG-7 edge histogram is designed to contain
only 80 bins describing the local edge distribution. These 80 histogram bins are the only
standardized semantics for the MPEG-7 EHD. However, using the local histogram bins only may
not be sufficient to represent global features of the edge distribution. Thus, to improve the retrieval
performance, we need global edge distribution as well.

20

Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.5(22)

Figure 5) Definition of sub-image and image-block

Figure 6) Five Edge Types

The EHD basically represents the distribution of 5 types of edges in each local area called a
sub-image. As shown in Fig. 5, the sub-image is defined by dividing the image space into 44 no
overlapping blocks. Thus, the image partition always yields 16 equal-sized sub-images regardless of
the size of the original image. To characterize the sub-image, we then generate a histogram of edge
distribution for each sub-image. Edges in the sub-images are categorized into 5 types: vertical,
horizontal, 45-degree diagonal, 135-degree diagonal and non-directional edges (Fig. 6). Thus, the
histograms for each sub-image represent the relative frequency of occurrence of the 5 types of
edges in the corresponding sub-image. As a result, as shown in Fig. 1, each local histogram contains
5 bins. Each bin corresponds to one of 5 edge types. Since there are 16 sub-images in the image, a
total of 516=80 histogram bins is required. Note that each of the 80-histogram bins has its own
semantics in terms of location and edge type. The semantics of the histogram bins form the
normative part of the MPEG-7 standard descriptor. Specifically, starting from the sub-image at (0,0)
and ending at (3,3), 16 sub-images are visited in the raster scan order and corresponding local
histogram bins are arranged accordingly. Within each sub image, the edge types are arranged in the
following order: vertical, horizontal, 45-degree diagonal, 135-degree diagonal, and non directional.
Table 1 summarizes the complete semantics for the EHD with 80 histogram bins. Of course, each
histogram bin value should be normalized and quantized. For normalization, the number of edge
occurrences for each bin is divided by the total number of image-blocks in the sub-image.
Histogram bins

Semantics

BinCounts[0]
Vertical edge of sub-image at (0,0)
BinCounts[1]
Horizontal edge of sub-image at (0,0)
BinCounts[2]
45 degree edge of sub-image at (0,0)
BinCounts[3]
135 degree edge of sub-image at (0,0)
BinCounts[4]
Non-directional edge of sub-image at (0,0)
BinCounts[5]
Vertical edge of sub-image at (0,1)
:
:
BinCounts[74]
Non-directional edge of sub-image at (3,2)
BinCounts[75]
Vertical edge of sub-image at (3,3)
BinCounts[76]
Horizontal edge of sub-image at (3,3)
BinCounts[77]
45 degree edge of sub-image at (3,3)
BinCounts[78]
135 degree edge of sub-image at (3,3)
BinCounts[79]
Non-directional edge of sub-image at (3,3)
Table-1) Semantics of Local Edge bins
Image block is a basic unit for extracting the edge information. That is, for each image-block,
we determine whether there is at least an edge and which edge is predominant. When an edge
exists, the predominant edge type among the 5 edge categories is also determined. Then, the
21

Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.5(22)

histogram value of the corresponding edge bin increases by one. Otherwise, for the monotone
region in the image, the image-block contains no edge. In this case, that particular image-block does
not contribute to any of the 5 edge bins. Consequently, each image-block is classified into one of
the 5 types of edge blocks or a non edge block. Although the non edge blocks do not contribute to
any histogram bins, each histogram bin value is normalized by the total number of image-blocks
including the non edge blocks. This implies that the summation of all histogram bin values for each
sub-image is less than or equal to 1. This in turn, implies that the information regarding non edge
distribution in the sub-image (smoothness) is also indirectly considered in the EHD. Now, the
normalized bin values are quantized for binary representation. Since most of the values are
concentrated within a small range (say, from 0 to 0.3), they are nonlinearly quantized to minimize
the overall number of bits Since the EHD describes the distribution of non-directional edges and
non edge cases as well as four directional edges, the edge extraction scheme should be based on the
image-block as a basic unit for edge extraction rather than on the pixel. That is, to extract
directional edge features, we need to define small square image-blocks in each sub-image as shown
in Fig. 5. Specifically, we divide the image space into non overlapping square image-blocks and
then extract the edge information from them. Note that, regardless of the image size, we divide the
image space into a fixed number of image-blocks . The purpose of fixing the number of imageblocks is to cope with the different sizes (resolutions) of the images. That is, by fixing the numberof-image blocks, the size of the image block becomes variable and is proportional to the size of the
whole image. The size of the image-block is assumed to be a multiple of 2. Thus, it is sometimes
necessary to ignore the outmost pixels in the image to satisfy that condition. Simple method to
extract an edge feature in the image block is to apply digital filters in the spatial domain. To this
end, we first divide the image-block into four sub-blocks. Then, by assigning labels for four subblocks from 0 to 3, we can represent the average gray levels for four sub-blocks at (i,j)th imageblock as a0(i,j), a1(i,j), a2(i,j), and a3(i,j), respectively. Also, we can rep-resent the filter
coefficients for vertical, horizontal, 45-degree diagonal, 135-degree diagonal, and non-directional
edges as fv(k), fh(k), fd-45(k), fd-135(k), and fnd(k), respectively, where k=0,12,,3 represents the
location of the sub-blocks. Now, the respective edge magnitudes mv(i,j), mh(i,j), md-45(i,j), md-135(i,j),
and mnd(i,j) for the (i,j)th image-block can be obtained as follows:

Figure 7) Sub-blocks and their labeling

22

Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.5(22)

Figure 8) Filter Coefficients for edge detection

The maximum value among 5 edge strengths obtained from (1) to (5) in the image-block is
considered to have the corresponding edge in it. Otherwise, the image-block contains no edge.
max{mv (i, j ), mh (i, j ), md 45 (i, j ), mnd (i, j )}

.....(6)

6.4 Distance Measures

We tried 3 kinds of histogram distance measures for a histogram H(i), i=1,2,.., N
6.4.1 L-2 Distance Defined as:
1

N
2
d (q, t ) = (hq (m) ht (m)) 2
m =1

this Metric is uniform in terms of the Euclidian distance between vectors in feature space, but
the vectors are not normalized to unit length (infact, they are on a hyperlane if the histogram is
normalized).
6.4.2 Cosine Distance
If we normalize all vectors to unit length, and look at the angle between them, we have cosine
distance, defined as:
N

hq ( m ) ht ( m )

2
d ( q , t ) = cos 1 m =1

min( hq , ht )

6.4.3 Histogram Intersection

Defined as:

The denominator term is needed for non-normalized histogram features

6.5 Combining Features and Making Decisions
This section formulates the problem of combining different features as a problem of finding a
set of weights of different feature distances, and then presents a Mini-Max algorithm in finding the
best-matching image,
23

Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.5(22)

Mini-Max Combination
Lets first look at a more general case, where we have: query image q, images in the database
i, K features and thus K kinds of distances (they are constants)
d k (q, i ), k = 1,.., K , i = 1,.., N

Assume we are going to combine them as a weighted sum of all the distances, i.e. the distance
for a image in the database is written as:
K

D ( q, i ) =

wk d k (q, i )

......(1)

and

k =1

= 1, wk 0, k = 1, 2,.., K

......(2)

k =1

Now we want to search for a vector w that satisfies Eq.(2) and the resulting distance measure
is "most close" to our subjective criteria. There are two candidate approaches:
i) assign a set of weights based on the perceptually judgment of the designer on some image
set (training). But the problem here is that this set of weights may perform poorly on new dataset.
ii) Or, having no assumption about the subjective judgment of a user, we choose the image
that minimizes the maximum distance over all valid set of weights as the best match(denoted as
Mini-Max hereafter). For every image i, searching for the maximum distance over the weight space
turns our to be a linear program, an thus have fast solution:
Maximize: (1), Subject to (2).Where all ds are the constants and wk , k = 1.., K are unknown. The
image with the miminum "max-distance" is declared as the best match to the query image. For our 2
features case the max distance
D(q, i ) = wdc (q, i ) + (1 w)de (q, i ), 0 w 1

of every image i, is a linear function of w over [0,1]. Thus the maximum either lies at w=0 or
w=1, and comparing dc (q, i) and de (q, i) is sufficient. Then we rank the maximum of
d c (q, i ) and d e (q, i ) for all i, and take n images with the least distance as our return result.
7 TESTING
7.1 Recall and Precision Evaluation:
Testing the effectiveness of the content based image Retrieval about testing how well the
CBIR can retrieve similar images to the query image and how well the system prevents the return
results that are not relevant to the source at all in the user point of view. A sample query image must
be selected from one of the image category in the database. When the system is run and the result
images are returned, the user needs to count how many images are returned and how many of the
returned images are similar to the query image. Determining whether or not two images are similar
is purely up to the users perception. Human perceptions can easily recognise the similarity between
two images although in some cases, different users can give different opinions. After images are
retrieved, the systems effectiveness needs to be determined. To achieve this, two evaluation
measures are used. The first measure is called Recall. It is a measure of the ability of a system to
present all relevant items. The equation for calculating recall is given below:
Re call =

Number of relevant items Re trieved

Number of relevant items in Collection

The second measure is called Precision. It is a measure of the ability of a system to present
only relevant items. The equation for calculating precision is given below.
Precision =

Number of relevant items retrieved

Total number of items retrieved

The number of relevant items retrieved is the number of the returned images that are similar
to the query image in this case. The number of relevant items in collection is the number of images
that are in the same particular category with the query image. The total number of items retrieved is
the number of images that are returned by the system.
24

Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.5(22)

7.2 Test cases

The CBIR implementation consists of various sub-systems for color and texture that are built
out of modules, which are composed of procedures and functions. The testing process hence
consists of different stages, where testing is carried out incrementally in conjunction with system
implementation. We have conducted several color and texture experiments on the test database of
250 images. The sample results of these experiments are given as follows:
7.2.1 Color test
Test

1
2
3
4
5

Total Number
of
Relevant
Images
46
38
18
30
32

Number of
relevant images
Retrieved

Total Number
of
Retrieved
Images
29
45
30
44
12
20
25
36
23
35
Table 2) Color Test cases

Recall

63.04
78.94
66.66
83.33
71.80

Precision

64.44
68.00
80.00
69.44
65.70

7.2.2 Texture Test

Test

1
2
3
4
5

Total Number
of
Relevant
Images
46
38
18
30
32

Number of
relevant images
Retrieved

Total Number
of
Retrieved
Images
19
45
28
44
11
20
22
36
19
35
Table 3) Texture Test cases

Recall

41.30
73.68
61.11
73.33
59.37

Precision

42.22
63.64
55.00
61.11
54.28

7.2.3 Color and Texture Test

Test

1
2
3
4
5

Total Number
Number of
Total Number
of
relevant images
of
Recall
Relevant
Retrieved
Retrieved
Images
Images
46
35
45
76.08
38
33
44
86.84
18
15
20
83.33
30
26
36
86.67
32
27
35
84.38
Table 4) Color and Texture Test cases

Precision

77.77
75.00
75.00
72.22
77.14

8 CONCLUSION
The dramatic rise in the sizes of images databases has stirred the development of effective and
efficient retrieval systems. The development of these systems started with retrieving images using
textual annotations but later introduced image retrieval based on content. This came to be known as
25

Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.5(22)

CBIR or Content Based Image Retrieval. Systems, using CBIR we can retrieve images based on
visual features such as colour, texture and shape, as opposed to depending on image descriptions or
textual indexing. In this paper we proposed an image retrieval system that evaluates the similarity of
each image in its data store to a query image in terms of color and textural characteristics, and
returns the images within a desired range of similarity. From among the existing approaches to
color and texture analysis within the domain of image processing, we have adopted the MPEG-7
edge histogram and color histogram to extract texture and color features from both the query images
and the images of the data store. For distance measuring between histogram vectors of two images,
we have tried three distance measures, they are l2 distance measure, cosine distance measure and
histogram intersection distance measure. The experiment results showing that, the average relevant
image retrieval rate is increased by 10% by combined features of color and texture than considering
features individually.

26

Georgian Electronic Scientific Journal: Computer Science and Telecommunications 2009|No.5(22)

9 REFERENCES
[1] Efficient use of MPEG-7 edge histogram descriptor- chee sun won, Dong kown park and soojun park ETRI Journal, volume 24, number 1, February 2002.
[2] Ak jain and a vailaya Image retrieval using color and shape, Pttern recogn vol 29 no 8, 1966
pp1233-1244
[3] Fundamentals of content based image retrieval Dr. fuhui long, Dr. Hongjiang and Prof. David
Dagan Feng
[4] Lars Jacob hove, Extending image retrivel systems with a thesaurus for shapes, Institute for
information and media sciences, University of Bergen
[5] Lu, G (1999) multimedia Database Management Systems. Norwood, Artech House INC
[6] Manjunath, B.S. and W.-y.Ma(2002) Texture Features for Image Retrieval. Image
Databases:Search and Retrieval of Digital Imagery V. castelli and R.Baeza-Yates, John
Wiley&sons, Inc:313-344
[7] K. Arbter, W. E. Snyder, H. Burkhardt, and G. Hirzinger, "Application of affine-invariant
Fourier descriptors to recognition of 3D objects," IEEE Trans. Pattern Analysis and
MachineIntelligence, vol. 12, pp. 640-647, 1990.
[8] E. M. Arkin, L.P. Chew, D..P. Huttenlocher, K. Kedem, and J.S.B. Mitchell, "An efficiently
computable metric for comparing polygonal shapes," IEEE Trans. Pattern Analysis and
Machine Intelligence, vol. 13, no. 3, pp. 209-226, 1991.
[9] J. Assfalg, A. D. Bimbo, and P. Pala, "Using multiple examples for content-based retrieval,"
Proc.Int'l Conf. Multimedia and Expo, 2000.
[10] J. R. Bach, C. Fuller, A. Gupta, A. Hampapur, B. Horowitz, R. Humphrey, R. Jain, and C. F.
Shu, "The virage image search engine: An open framework for image management," In Proc.
SPIC Storage and Retrieval for Image and Video Database, Feb. 1996.
[11] N. Beckmann, et al, "The R*-tree: An efficient robust access method for points and
rectangles," ACM SIGMOD Int. Conf. on Management of Data, Atlantic City, May 1990.
[12] A. Blaser, Database Techniques for Pictorial Applications, Lecture Notes in Computer
Science,Vol.81, Springer Verlag GmbH, 1979.
[13] P. Brodatz, "Textures: A photographic album for artists & designers," Dover, N Y, 1966.
[14] H. Burkhardt, and S. Siggelkow, "Invariant features for discriminating between equivalence
classes," Nonlinear Model-based Image Video Processing and Analysis, John Wiley and Sons,
2000.
[15] C. Carson, M. Thomas, S. Belongie, J. M. Hellerstein, and J. Malik, "Blobworld: A system for
region-based image indexing and retrieval," In D. P. Huijsmans and A. W. M. Smeulders, ed. Vi s
u al Information and Information System, Proceedings of the Third International
Conference VISUAL99, Amsterdam, The Netherlands, June 1999, Lecture Notes in Computer
Science 1614.Springer, 1999.
[16] J.A. Catalan, and J.S. Jin, "Dimension reduction of texture features for image retrieval using
hybrid associative neural networks," IEEE International Conference on Multimedia and Expo,
Vol.2, pp. 1211 -1214, 2000.
[I7]A. E. Cawkill, "The British Library's Picture Research Projects: Image, Word, and
Retrieval,"Advanced Imaging,Vol.8, No.10, pp.38-40, October 1993.
[18] N. S. Chang, and K. S. Fu, "A relational database system for images," Technical Report TREE 79-82, Purdue University, May 1979.

_________________________
Article received: 2008-12-18

27