International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 05 Issue: 04 | Apr-2018 www.irjet.net p-ISSN: 2395-0072
© 2018, IRJET | Impact Factor value: 6.171 | ISO 9001:2008 Certified Journal | Page 2425
A REVIEW ON CONTENT BASED IMAGE RETRIEVAL BASED ON SHAPE,
COLOR AND TEXTURE FEATURES USING DWT, MODIFIED K-MEANS
AND ANN
Manisha Aeri
1
, Ashok Kumar
2
, H.L. Mandoria
3
, Rajesh Singh
4
1
M.Tech Student, Department of Information Technology, GBPUAT Pantnagar, Uttarakhand, India
2,4
Asst. Professor, Department of Information Technology, GBPUAT Pantnagar, Uttarakhand, India
3
Professor and Head, Department of Information Technology, GBPUAT Pantnagar, Uttarakhand, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Content based image retrieval (CBIR), as the
name suggests is the retrieval of images based on some of the
visual features like color, shape, texture etc. It has proved
being a champion among the most remarkable research areas
in the recent years and it's need can be found in various
specific domains, for example, Data processing, Education,
Medical Imaging, Crime bar, Weather surveying etc. This
paper reviews the completely unique technique that uses an
effective calculation for Content Based Image Retrieval (CBIR)
in context of Discrete Wavelet Transform (DWT), Modified K-
Means Clustering and Artificial Neural Network. There are two
basic steps to be followed in CBIR i.e. feature extraction and
similarity measurement. This paper comparatively utilizes
wavelet transform which helps in the image compression and
denoising. Image compression helps to reduce the storage
space of images which can eventually increase the
performance. Discrete wavelet transformation decreases the
size of feature vector as well as preserve the content
details.ANN is more effective and efficient algorithm for the
similarity measurement and also ANN is used to train and test
the proposed framework. The blend of DWT, Modified K-Means
procedures and Neural Network expands the execution of
image retrieval structure for shape, shading and surface based
request. Trial happens demonstrate that the proposed plot has
higher retrieval exactness than other traditional plans like
Precision and Recall.
Key Words: dwt, modified k means, ANN, Gabor filter,
Haar wavelet
1. INTRODUCTION
In early days as a result of extensive image accumulations the
manual approach was more tough so as to beat these
difficulties Content Based Image Retrieval (CBIR) was
presented. Content-based image retrieval (CBIR) is the use of
laptop vision to the image retrieval drawbacks. During this
approach rather than being physically annotated by textual
keywords in this approach pictures would be indexed
utilizing their own particular visual contents .The visual
contents might be shading, surface and shape. This approach
is alleged to be a general framework of picture retrieval.
There are three key bases for Content Based Image Retrieval
which is visual component extraction, multidimensional
categorization and retrieval system design. The shading
viewpoint can be accomplished by the strategies like
averaging and histograms. The surface angle can be
accomplished by utilizing changes or vector quantization .The
shape viewpoint can be accomplished by utilizing gradient
operators or morphological operators.
A picture retrieval system may be a system that permits us to
browse, to make a search and retrieve the pictures. Content
Based Image Retrieval is that the method of retrieving the
required question image from a large range of databases that
rely on the contents of the image. Color, texture, shapes and
other native features are the square measures or the
overall techniques used for retrieving a selected image
from the pictures database. Content based mostly Image
Retrieval systems works with all the pictures and therefore
the search is based on comparison of features with the
question image.
The principle elements of CBIR are the features which
incorporates the Geometric shapes, hues and the texture of
the picture. There are basically two types of features that are
global and local features. Object recognition should be
possible effortlessly by utilizing the local features. The
consequent element is the related content or text in which
the pictures can likewise be recovered utilizing the content or
text related with the picture. The other element is
the relevant feedback wherever it helps to be a lot more
precise in making the search of relevant pictures just
by absorbing the feedbacks of the user.
1.1 Feature Extraction
Feature extraction is a methodology that is applied to any
image so that we can categorize and recognize the pictures
from huge set of data on the basis of those features. The
features can be color, shape, texture etc.
1.2 Color
One of the most beneficial and distinguishing feature is the
color. A color histogram methodology is more effective and
efficient, that’s why it is more frequently used for CB)R. For
more color histogram match HSV color space are used. The
use of HSV color space is to manipulate the hue and
saturation.
1.3 Color Histogram
If the color pattern is unique and that color pattern is
compared with the massive range of the data set in that case