2012 IEE International Conference on Advanced Communication Control and Computing Technologies (ICACCCT)
An Effective Content Based Image Retrieval (CBIR) System Based on
Evolutionary Programming (EP)
*Sunita Manoj Jadhav,**Dr.Vikram Patil
*Assistant Professor Department ofEXTC Saaswati college of Engineering Kharghar, Navi Mumbai.
**Principal SanjeevanEngineering and Technology Institute Panhala, Kolhapur.
*sunitaanoladhavphd@gmail.com, * *vsp.research@gmail.com
Abstract- This paper introduces a content Based Image
Retrieval (CBm) based on evolutionary algorithm.
Initially, the shape, color and texture feature is extracted
for the given query image and also for the of the database
images in a similar manner. Subsequently, similar images
are retrieved utilizing an evolutionary algorithm based
similarity. Thus, by means of the evolutionary algorithm,
the required relevant images are retrieved from a large
database based on the given query. The proposed CBIR
system is evaluated by querying different images and the
efciency of the proposed system is evaluated by means of
the precision-recall value of the retrieved results.
Kewords- feature etraction, shape, color histogram
teure, query, Evolutionar Pogramming (EP)
I. INTRODUCTION
Content Based Image Retrieval (CBIR) is an
important research area for manipulating large
multimedia databases and digital libraries [1] [2]
characterized by automatic indexing of images based on
their own visual features [7] [14] [15]. Commonly used
features include [3] color, texture, shape, and edge
information. Many retrieval methods that accept query
images as input fom the user represent images as
vectors in the feature space and search for images based
on their features and feature representations [13]. When
the user presents a sample query image, region of
interest (ROI), or pattern to the system, it performs
various visual query mechanisms, such as the query-by
example (QBE) paradig, and fnally outputs the
relevant images [6]. In CBIR, image content is
fequently represented using image features [5].
CBI fnds applications in internet, advertising,
medicine, crime detection, entertainment, and digital
libraries. High retieval efciency and less
computational complexity are the desired characteristics
of CBI system and they are the key objectives in the
desig of a CBIR system [4]. Many CBI systems have
been proposed in the past decade, including QBIC,
Virage, Photohook, Visual SEEk, Netra, MARS, and so
on. Unfortunately, there are still many problems that
hinder CBI systems fom being popular [9] [10]. A
CBI system should efectively capture the user's
preference on retrieval by well characterizing a mapping
fom image features to human concepts [S]. Two major
research communities that are concerned with image
retrieval are database management and computer vision.
While one of these two research communities deals with
text-based image retrieval, the other is associated with
ISBN No. 97S-1-4673-204S-1112/$31.00©2012 IEEE
visual-based image retrieval [11]. However, there
exist two major diffculties, especially when the size of
the image collection is large (tens or hundreds of
thousands). One is the large amount of labor required in
manual image annotation, whereas the other one, which
is more important, results fom the rich content in the
images and the subjectivity of human perception [12].
Traditionally, the majority of CBI systems utilize
any one or two feature of the image for analyzing the
similarity and due to this the efectiveness of the CBI
system is degraded. Hence to overcome this, in this
paper we extract an extensive feature set and for
retrieving the relevant images we utilize the
evolutionary programming method. Te color, shape
and texture features are exracted in the feature
extraction process. Consequently, evolutionary
programming is applied to obtain the relevant images
for the given query image. Te rest of the paper is
organized as follows section 2 reviews the recent
research works related to CBIR techniques. Te steps
involved in our proposed work with mathematical
formulation and pictorial representation are detailed in
section 3. Section 4 discusses about the simulation
results. Section 5 concludes the paper.
II. RECENT RELATED RE SEARCHES: A REVIEW
A handfl of recent researches that are related to the
proposed concept are discussed in detail in this section.
Hiremath et al. [16] have presented a method for salient
points determination based on color saliency. The color
and texture information around these points of interest
have served as the local descriptors of the image. In
addition, the shape information has been captured in
terms of edge images computed using Gradient Vector
Flow felds. Invariant moments have been used to
record the shape features. The combination of the local
color, texure and the global shape features has
provided feature set for image retrieval.
Arti Khaparde et af [17] have presented an
approach for global feature extraction using
Independent Component Analysis (ICA). A
comparative study has been done between ICA feature
vectors and Gabor feature vectors for ISO different
texture and natural images in a databank. Result
analysis has shown that extracting color and texture
information by ICA has provided sigifcantly
improved results in terms of retrieval accuracy,
computational complexity and storage space of feature
310