CBIR Processing Approach on Colored and Texture Images using KNN Classifier and Log-Gabor Respectively

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 04 Issue: 06 | June -2017 www.irjet.net p-ISSN: 2395-0072
© 2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 546
CBIR Processing Approach on Colored and Texture Images using KNN
Classifier and Log-Gabor Respectively
Arti, Astt.Prof. Nancy
Department of CSE, International Institute of Engineering and Technology (Samani,KKR) India
------------------------------------------------***-------------------------------------------------
Abstract: Content Based Image Retrieval (CBIR), also
called as Query By Image Content (QBIC). Content Based
Image Retrieval is the method to retrieve stored image
from database by supplying query image instead of text.
This is achieved using proper feature extraction and
matching process. Here we have implemented two
methods of content based image retrieval using color and
texture. In feature extraction of color is done using
classifiers and similarity measure,color moment. While
feature extraction of texture is done using wavelet texture
features and Log-Gabor features. Finally we have retrieved
top images using euclidean distance and chisquare
distance and we have made comparative analysis. Content
Based Image Retrieval has endless discussion to do. Here
we can say that results or retrieval ratio depends upon
image class for some images. we can have better precision
and time complexity while some images give average
result. Finally comparative analysis given in table 1 says
that overall precision and time complexity given by
combined approach using classifier, similarity measurev
and log-gabor respectively color and texture gives better
result as compared to wavelets and gabor filter. Different
types of classification we can use Neural network, Support
Vector Machine (SVM), KNN, Bayesian etc. In this paper,
we are using K Nearest Neighbor (KNN) classifier to find
out the relevant images and after that we use Spearman’s
Rank Correlation Function to reduce the time complexity
and improve F-measure. Hence if we want to improve
retrieval efficiency we have to use some other approach.
Here for effective retrieval we can use other features like
shape.
Keywords: CBIRS, Image databases, Color string
comparison, Feature extraction, Query image, Target
Image.
I. INTRODUCTION
With the headway in internet and multimedia
technologies, a immense amount of multimedia data
in the form of audio, video and images has been utilized as
a part of numerous fields like medical treatment, satellite
data, video and still images repositories and surveillance
system. This has made a progressing interest of
frameworks that can store and retrieve mixed media
information in a powerful way.
Numerous multimedia information storage and retrieval
systems have been developed till now to cater these
requests.
The most common retrieval systems are Text Based Image
Retrieval (TBIR) systems, where search is based on
automatic or manual explanation of images. A
conventional TBIR searches database for the similar text
surrounding the image as given in the query string. The
TBIR systems are fast as the string matching is
computationally less time consuming process. In addition
annotation of images is not always correct and takes a lot
of time. For finding alternative way of looking and
overcoming the limitations forced by TBIR systems more
natural and easy to understand content based image
retrieval systems (CBIR) were developed. A CBIR system
uses visual contents of images described in the form of low
level features like color, texture, shape and spatial
locations to represent images in databases. The system
retrieves comparable images when an query image or
sketch is presented as input to the system. Querying in this
way removes need for describing the visual content of
images in words and is near to human perception of visual
data. A portion of representative CBIR systems is Query by
Image Content (QBIC).
Fig.(1) Block Diagram of CBIR
In a typical CBIR system (Figure 1.2), image low level
features like color, texture, shape and spatial locations are
represented in the form of a multidimensional feature
vector. The query image is converted into the internal
representation of feature vector using the same feature
extraction routine that was used for building the feature

database. The similarity measure is employed to calculate
the distance between the feature vectors of query image
and those of the target images in the feature database.
Finally, the retrieval is performed using an indexing
scheme which facilitates the efficient searching of the
image database. Recently, user’s relevance feedback is also
incorporated to further improve the retrieval process in
order to produce perceptually and semantically more
meaningful retrieval results. In this, we discuss these
fundamental techniques for content- based image
retrieval.
II. RELATED WORK
M. Kaipravan and Rejiram R, et al [28] In this paper, we
present a CBIR system based on integration of both color
and texture feature. Due to the poor discriminating power
of color histogram, color moments that encode some
spatial information are used to extract the color feature
from the image. Gabor filter is used to represent the
texture feature. Then we assign weights to each feature
and calculate the similarity of combined features using
Manhattan distance measure. we have proposed an
efficient image retrieval system based on combination of
color moments and Gabor texture feature. Gabor filter are
adopted to extract texture feature and feature vector of
length 48 is obtained. We calculate the similarity with
combined features of color and texture using Manhattan
distance as the similarity measure. The proposed method
has higher retrieval accuracy than other conventional
methods.
P. B. Patil and M. B. Kokare, et al [12] Due to the semantic
gap between low-level image features and high level
concepts, we have presented a framework for effective
image retrieval by proposing a novel idea of cumulative
learning using Support Vector Machines (SVM). It creates a
knowledge base model to increase the training samples by
simply accumulating the samples based on user
interactions. As we know relevance feedback (RF) is online
process, so we have optimized the learning process by
considering the most positive image selection on each
feedback iteration. To learn the system we have used SVM.
The main significances of our system are to address the
small training sample and to reduce retrieval time. In this
paper, an active relevance feedback framework has been
proposed to handle the small training data problem in RF
and optimizing the testing set in order to reduce the
retrieval time. The proposed relevance feedback
framework, we found that RF using SVM with combined
texture features RCWF and DT-CWT gives better retrieval
performance than contourlet and curvelet texture features.
R. Sudhakar, K. R. Krishnan and S. Muthukrishnan, et al[3]
with an increase in the awareness of internet usage, there
has been an explosion of data on the web. The problem of
retrieving near approximate images using textual queries
has always been an area of research. This focuses on
bridging the gap between textual search input given by the
user and the images retrieved from the database, by
making use of visual features instead of the file name.
The work concentrates on employing a simple keyword
extraction technique rather than using complex NLP
techniques. Also, an automatic segmentation method is
proposed to avoid human intervention. It refining the
segmentation algorithm and improving the template
matching techniques would further improve the retrieval
efficiency. the spot and eye detection methods can be
applied on medical image datasets to identify circular
portions such as cysts, tumors or organs. This paper
considers a single isolated object.
Roshi Choudhary et al. [7] proposed an approach to
perform content based image retrieval. It is an integrated
approach used to extract color and texture feature from
images. By using single feature, correct results can never
produced. So multi feature extraction is more beneficial to
perform image retrieval. To extract the color feature,
higher order of color moment is used which is the
descriptor of color. To extract texture, LBP is used which is
the descriptor of texture. Local binary pattern is mainly
used to face recognition.
Vinee. V. Kawade et al. [10] announced a user based system
for CBIR in which genetic algorithm is applied. The
different features of color image like mean, standard
deviation and the image bitmap are used for retrieval. The
texture features like edge histogram of an image and the
entropy of gray level co-occurrence matrix are used.
Moreover, the genetic algorithm is applied to help user in
identifying the images which satisfy his needs for reducing
gap between the users’ expectation and the retrieval
results. Experimental results show remarkable
improvement in the performance after applying IGA.
L. Chai, H. Zhang, Z. Qin, J. Yu and Y. Qi, et al[4] Content-
based image retrieval (CBIR) has got an intense interest
and seen considerable progress over the last decade. But
most of the time it is only applied in laboratory. One
important reason for this is the diversity of images.. At
present, and even in the foreseeable future, a general
purpose CBIR system is not really possible. In this paper,
we propose a region-based method fit for the content-
based retrieval of product images. The method focuses on
two key issues: fast extraction of the main region, in which
the product locates, as well as efficient shape and color
features extraction. To show the validity of the proposed
region-based method, compared experiments are carried
out and illustrated on the PI 100 dataset.
S. Selvarajah and S. R. Kodithuwakku, et al [20]
Representation of visual features and similarity match are

important issues in CBIR. Colour and texture features are
important properties in CBIR systems. In this paper, a
combined feature descriptor for CBIR is proposed to
enhance the retrieval performance for CBIR. This method
is developed by exploiting the wavelets and colour
histogram moments. First, Haar wavelet is used to
decompose colour images into wavelet coefficients. Then
Second image feature extraction and similarity matching
are performed by means of histogram moments.
Kommineni Jenni et al. [8] presented a Content Based
Image Retrieval approach based on the database
classification using Support Vector Machine (SVM) and
color string coding feature selection. In SVM method, the
feature extraction was done based on the basis of color
string coding and string comparison. Here, they succeed in
transferring the images retrieval problem to strings
comparison. Thus the computational complexity is
decreases obviously and increased the accuracy in
obtaining results for image retrieval. Using database
classification we can improve the performance of the
content based image retrieval.
Ammar Huneiti et al. [9] proposed a CBIR method by
extracting both color and texture feature vectors using the
Discrete Wavelet Transform (DWT) and the Self
Organizing Map (SOM) artificial neural networks. At query
time texture vectors are compared using a similarity
measure which is the Euclidean distance and the most
similar image is retrieved. In addition, other relevant
images are also retrieved using the neighborhood of the
most similar image from the clustered data set via SOM.
Results showed that the proposed method is able to
retrieve images with higher average precision values than
other methods proposed in literature by just comparing
the texture similarity and without any need to compare
color similarities.
S. Selvarajah and S. R. Kodithuwakku, et al [13]
Representation of visual features and similarity match are
important issues in CBIR. Color and texture features are
important properties in CBIR systems. In this paper, a
combined feature descriptor for CBIR is proposed to
enhance the retrieval performance for CBIR. This method
is developed by exploiting the wavelets and colour
histogram moments. First, Haar wavelet is used to
decompose colour images into wavelet coefficients. Then
image feature extraction and similarity matching are
performed by means of histogram moments.
Siddarth Ladhake et al. [6] provides a system for large
scale database is designed and implemented. Here,
proposed system exploits semantic binary code
generation techniques, fine and coarse similarity measure
technique, which improves accuracy, image retrieval
speed. Finally,the performance of image retrieval is
improved in terms of accuracy, retrieval time and
efficiency.
Devyani Soni et al. [14] proposed an efficient color space
Based Approach for Image Retrieval Using fusion of Color
Histogram and color correlogram. During experimentation,
both HSV color model as well as RGB color model was used
for the same process of retrieval and it was observed that
HSV color space gives more accurate result as compared to
RGB color space.
Priyadarshini Patil et al. [2] proposed and implemented
an efficient image retrieval technique using both color and
texture features of an image. Here they compare and
analyze performance of an
image retrieval using both these features. And we see
that CBIR using color features gives high precision where
as CBIR using texture feature features give high recall.
III. PROPOSED OF CBIR
Content Based Image Retrieval is a technique that enables
a user to extract an image based on a query from the
database containing huge amount of images. Here, we have
to test a query image from our own built Dataset and
provides the accurate result to the user.
IV. PROPOSED WORK
In this paper, we are going to propose Content based
Image Retrieval System on the colored images. Here,
we built our own dataset. The Dataset is divided into
two categories: Train Data and Test Data. The steps
involved are:
 The Query image is given by the user
corresponding to which user want the results.
 Read that particular Query image.
 Extract features from the Query image on the
basis of color and relatively find the Prediction
class using K Nearest Neighbor (KNN) classifier.
Prediction class is used to find out the relevant
images.
 Now to order these relevant images we use
Spearman’s Rank Correlation Function to
calculate the distances of each relevant image with
Query image. It will sort all these relevant images
and fetch top-n images (top-n < size of Dataset)
and print these images.
K Nearest Neighbor: K nearest neighbors is a simple
algorithm that stores all available cases and
classifies new cases based on a similarity measure
(e.g., distance functions). KNN has been used in
statistical estimation and pattern recognition.
Spearman's rank correlation coefficient or Spearman's
rho: is a nonparametric measure of rank correlation

(statistical dependence between the ranking of two
variables). It assesses how well the relationship between
two variables can be described using a monotonic function.
Mathematical notation of Spearman Rank formula is
The result will
always be between 1 and minus 1.
Here,
d= Distances
n= No. of cases
Log Gabor: Field introduced the Log-Gabor filter
and showed that it is able to better encode natural images
compared with the original Gabor filter. Additionally, the
Log-Gabor filter does not have the same DC problem as the
original Gabor filter. A one dimensional Log-Gabor
function has the frequency response:
V. EXPERIMENT AND RESULTS SYSTEM DETAILS
A. Hardware details
We have validated our results on machine with the
configuration of installed memory (RAM 3GB), 64-
bit Operating System, having processor Intel(R)
Core(TM) i3-2310M CPU @ 2.10GHz. Here, we
have our own created Dataset e.g., images of Flags of
different countries.
B. Software details
MATLAB 7.0
WINDOW 7
The Experimental work is done on the MATLAB. MATLAB
is a software package for high performance numerical
computation and visualization. It provides an interactive
environment with hundreds of built-in functions for
technical computation, graphics and animation. The name
MATLAB stands for MATrix LABoratory. MATLAB is an
efficient program for vector and matrix data processing. It
contains ready functions for matrix manipulations and
image visualization.
MATLAB provides a suitable environment for image
processing. Although MATLAB is slower than some
languages (such as C), its built in functions and syntax
makes it a more versatile and faster
programming environment for image processing. In this
paper, proposed work is done on MATLAB as it contains
ready-made functions so this tool is easy to use. We are
going to compare existing and proposed approach.
A precision rate can be defined as the number of
relevant images retrieved by a search divided by the
total number of images retrieved by that search. The
equation is as follows:
Precision =
Where A is relevant correctly retrieved and B is
falsely retrieved.
A recall rate is defined as the number of relevant images
retrieved by a search divided by the total
number of existing relevant images (which should have
been retrieved). The equation is as follows have been
retrieved). The equation is as follows.
Table1 : Comparison of Existing and Proposed approach
Approach Precision Recall F-Measure Time
complexity
Existing 0.625 0.625 0.625 1.62
Proposed 1 1 1 1.34

The validated proposed results are shown below
from figure 1(a,b)
Figure 1(a): Query image of Rose to be searched
Figure 1(b): Results corresponding to the Query Image
of Rose
The results of Texture Base Proposed approach
is shown below
in figure 2(a,b)
Figure 2(a): Query Image for Texture base feature
Figure 2(a): Query Image for Texture base feature

These results can also be explained with the help of graphs:
Here Graph1,2 and 3 shows the F-Measure and Time Complexity and Accuracy respectively
Graph 1: Comparison of Existing and Proposed Approaches on the basis of F-measure.
Graph 2: Comparison of Existing and Proposed Approaches in terms of Time Complexity.
Graph 3: Comparison of Existing and Proposed Approaches in terms of Accuracy

VI. CONCLUSION AND FUTURE SCOPE
From above discussion we conclude that the proposed approach perform better than existing approach with reduced time
complexity and improves F-measure value. Also we use Nearest Neighbor (KNN) Classification to calculate the relevant images
from Dataset and Spearman’s Rank Correlation Function for calculating the distances and Log-Gabor feature of texture. Gabor
filters have not zero mean, which produces a non-uniform coverage of the Fourier domain. This distortion causes fairly
poor pattern retrieval accuracy. To address this issue, we propose a simple yet efficient image retrieval approach based on
a novel log-Gabor filter scheme. We make emphasis on the filter design to preserve the relationship with receptive fields
and take advantage of their strong orientation selectivity. As here we have own Dataset which contain limited number of
images. In future, we can use this concept for the huge database as well and use some other classifiers to enhance the results.
REFRENCES
[1] Priyadarshini Patil and Bhagya Sunag, Analysis of
Image Retrieval Techniques Based On Content, IEEE
International Advance Computing Conference (IACC).
[2]. R. Sudhakar, K. R. Krishnan and S. Muthukrishnan, "A
hybrid approach to content based image retrieval using
visual features and textual queries," 2011 Third
International Conference on Advanced Computing,
Chennai, 2011, pp. 241-247.
[3]. L. Chai, H. Zhang, Z. Qin, J. Yu and Y. Qi, "Multi-feature
content-based product image retrieval based on region of
main object," 2011 8th International Conference on
Information, Communications & Signal Processing,
Singapore, 2011,pp.1-5.
[4] Anuja khodaskar, Siddarth Ladhake, Promising Large
Scale Image Retrieval by using Intelligent Semantic Binary
Code Generation Technique, International Conference on
Intelligent Computing Communication & Covergence
(ICCC-2014) Conference Organized by Interscience
Institute of Management and Technology, Bhubaneswar,
Odisha, India.
[5] Roshi Choudhary, Nikita Raina, Neeshu Chaudhary,
Rashmi Chauhan, Dr. R H Goudar, An Integrated Approach
to Content Based Image Retrieval, International
Conference on Advances in
Computing, Communications and informatics (lCACC1)
2014
[6] Kommineni Jenni, Satria Mandala, Mohd Shahrizal Sunar,
Content Based Image Retrieval Using Colour Strings Comparison,
2nd International Symposium on Big Data and Cloud Computing
(ISBCC’15)
[7] Ammar Huneiti, Maisa Daoud, Content-Based Image
Retrieval Using SOM and DWT, Journal of Software
Engineering and Applications, 2015, 8, 51-61 Published
Online February 2015
[8] Vinee. V. Kawade and Arti. V. Bang, Content Based
Image Retrieval Using Interactive Genetic Algorithm,
Annual IEEE India Conference (INDICON) 2014
[9]. P. B. Patil and M. B. Kokare, "Interactive content-based
texture image retrieval," 2011 2nd International
Conference on Computer and Communication Technology
(ICCCT-2011), Allahabad,2011,pp.71-76.
[10]. S. Selvarajah and S. R. Kodithuwakku, "Combined
feature descriptor for Content based Image Retrieval,"
2011 6th International Conference on Industrial and
Information Systems, Kandy, 2011, pp.164-168.
[11] Devyani Soni , K. J. Mathai, An Efficient Content Based
Image Retrieval System based on Color Space Approach
Using Color Histogram and Color Correlogram, Fifth
International Conference on Communication Systems and
Network Technologies 2015.
[12]. S. Selvarajah and S. R. Kodithuwakku, "Combined
feature descriptor for Content based Image Retrieval,"
2011 6th International Conference on Industrial and
Information Systems, Kandy, 2011, pp.164-168.
[13]. M. Kaipravan and Rejiram R, "A novel CBIR system
based on combination of color moment and Gabor filter,"
2016 International Conference on Data Mining and
Advanced Computing (SAPIENCE),
Ernakulam,2016,pp.170-174

CBIR Processing Approach on Colored and Texture Images using KNN Classifier and Log-Gabor Respectively

More Related Content

What's hot (20)

Similar to CBIR Processing Approach on Colored and Texture Images using KNN Classifier and Log-Gabor Respectively (20)

More from IRJET Journal (20)

Recently uploaded (20)

CBIR Processing Approach on Colored and Texture Images using KNN Classifier and Log-Gabor Respectively