2013 International Conference on Advanced Computing and Communication Systems (ICACCS -2013), Dec. 19 – 21, 2013, Coimbatore, INDIA
Content Based Image Retrieval System using Texture
and Modified Block Truncation Coding
Purohit Shrinivasacharya
Department of Information Science and Engineering
Siddaganga Institute of Technology
Tumkur, India
purohitsn@gmail.com
Dr. M. V Sudhamani
Department of Information Science and Engineering
R. N. S Institute of Technology
Bengalore, India
mvsudha_raj@hotmail.com
Abstract—According to the recent literature, it is pragmatic
that image storage and retrieval through the World Wide Web
(WWW) has made impressive progress. Practical searching for
image still confronts us with present retrieval systems. A Content
Based Image Retrieval (CBIR) system provides an efficient way
of retrieving related images from image collections. In this paper
we present a new technique to extract the images from the
database to achieve better performance and efficiency. The novel
method uses an approach, which combines texture information
and modified Block Truncation Coding (BTC) method to extract
the color features from the image. The whole image is divided
into a fixed block of size 4 x 4, 8 x 8, 16 x 16, 32 x 32, 64 x 64, 128
x 128 and 256 x 256 non overlapped blocks. For each block, we
calculated the threshold as a mean of the pixel values and the
number of pixels which are lower than threshold value and upper
than the threshold value are counted and mean of these lower
and upper values have been taken and these two are considered
as features. Now this process is repeated for the all blocks to
extract the features of the image. The second step will compute
the four texture features using the Gray-Level Co-occurrence
Matrix (GLCM) as a statistical approach to extract energy,
contrast, entropy and uniformity. We have used this modified
BTC and texture features of the image to retrieve images.
Finally, we have conducted experiments on a ground truth image
database that has 1000 images of different categories. A different
block sizes have been used to demonstrate the experimental
results and compared with tabulation. From this we have
observed that the performance is better and efficiency is low with
smaller block sizes and vice versa with higher block sizes. So we
have to choose the middle ranges of block sizes for better
performance with better efficiency. As we have observed from
the results that the performance of the retrieval system is
improved because of combination of texture and color features.
The proposed system’s performance has been improved by an
average precision of 2.41% and recall of 8.22% when compared
to existing systems.
Keywords—BTC; energy; database; color; CBIR; speed
I. INTRODUCTION
The amount of image data is growing very fast in the
present day, because of huge bandwidth and large volume of
storage for the little price. This creates a problem of retrieving
the stored image in the system, because extracting the image
with the surrounding textual information is very difficult and it
won't give the best result. Hence the present word is required
the different method to extract the stored images i.e CBIR. The
recent CBIR research tries to combine both of the Text Based
Image Retrieval and CBIR methods and has developed an
efficient image representation and data structures, query
processing algorithms, intelligent query interfaces and domain
independent system architecture [1]. Implementing the CBIR
system is very difficult because extracting the visual property
of an image is a tedious job. Most of the researchers are finding
the methods to extract the good visual features from the images
that have to take less time and less space to store this image
features.
The major aim of any image retrieval system is to retrieve
as many images as possible from an image database such that
these retrieved images meet the user’s requirements. The user’s
can be specified the requirements in the form of an image or a
sketch or a major color or keywords depending on type of
system. An image retrieval system provides the user with a
way to access, browse and retrieve efficiently and possibly in
real time, form these databases [2]. Image retrieval systems can
be divided into two main types: Text Based Image Retrieval
and Content Based Image Retrieval. In the early years Text
Based Image Retrieval was popular, but nowadays Content
Based Image Retrieval has been a topic of intensive research
[3] [4]. In this paper we are presenting a CBIR technique to
retrieve the images from the local database or from WWW.
The proposed method uses the texture analyzed [5] and gray
color information for extracting the features of the image for
later comparison. The texture information energy, contrast,
entropy and uniformity are extracted from the Gray-Level Co-
occurrence Matrix (GLCM) [6, 7] statistical approaches. The
gray color features are extracted by using the modified BTC
[8] method. The different category images are chosen for
experiment to test the proposed system performance. The Fig.
1. Shows the different categories of images used.
[978-1-4799-3506-2/13/$31.00 ©2013 IEEE]
Fig. 2. Proposed CBIR System