2010 IEEE International Conference on Computational Intelligence and Computing Research
ISBN: 97881 8371 362 7
Image Clustering and Retrieval using Image Mining Techniques
A.Kannan
1
Dr.V.Mohan
2
Dr.N.Anbazhagan
3
1
Associate Professor, Department of MCA, K.L.N. College of Engineering,
Sivagangai District, Tamilnadu, India – 630611
kannamca@yahoo.com
2
Professor & Head, Department of Mathematics, Thiagarajar College of Engineering,
Madurai,Tamilnadu,India - 625 015
vmohan@tce.edu
3
Reader, Department of Maths, Alagappa University, Karaikudi-04, Tamilnadu,India,
anbazhagan_n@yahoo.co.in
Abstract: Image retrieval is the basic requirement task
in the present scenario. Content Based Image Retrieval
is the popular image retrieval system by which the
target image to be retrieved based on the useful features
of the given image. In other end, image mining is the
arising concept which can be used to extract potential
information from the general collection of images.
Target or close Images can be retrieved in a little fast if
it is clustered in a right manner. In this paper, the
concepts of CBIR and Image mining have been
combined and a new clustering technique has been
introduced in order to increase the speed of the image
retrieval system.
KEY WORDS: Content Based Image Retrieval, RGB
Components, Texture, Entropy.
1.0 INTRODUCTION
In this present scenario, image plays vital role in
every aspect of business such as business images,
satellite images, medical images and so on. If we
analysis these data, which can reveal useful
information to the human users. But, unfortunately
there are certain difficulties to gather those data in a
right way [1]. Due to incomplete data, the
information gathered is not processed further for any
conclusion.
In another end, Image retrieval is the fast growing
and challenging research area with regard to both still
and moving images. Many Content Based Image
Retrieval (CBIR) system prototypes have been
proposed and few are used as commercial systems.
CBIR aims at searching image databases for specific
images that are similar to a given query image. It also
focuses at developing new techniques that support
effective searching and browsing of large digital
image libraries based on automatically derived
imagery features. It is a rapidly expanding research
area situated at the intersection of databases,
information retrieval, and computer vision. Although
CBIR is still immature, there has been abundance of
prior work.
The CBIR focuses on Image ‘features’ to enable the
query and have been the recent focus of studies of
image databases. The features further can be
classified as low-level and high-level features. Users
can query example images based on these features
such as texture, colour, shape, region and others. By
similarity comparison the target image from the
image repository is retrieved. Meanwhile, the next
important phase today is focused on clustering
techniques. Clustering algorithms can offer superior
organization of multidimensional data for effective
retrieval. Clustering algorithms allow a nearest-
neighbour search to be efficiently performed.
Hence, the image mining is rapidly gaining more
attention among the researchers in the field of data
mining, information retrieval and multimedia
databases. Spatial Databases is the one of the
concepts which plays a major role in Multimedia
System. Researches can extract semantically
meaningful information from image data are
increasingly in demand.
1.1 Comparison of Image Mining with other
Techniques
Image mining normally deals with the extraction of
implicit knowledge, image data relationship, or other
2010 IEEE International Conference on Computational Intelligence and Computing Research
patters not explicitly stored from the low-level
computer vision and image processing techniques.
i.e.) the focus of image mining is the in the extraction
of patterns from a large collection of images, the
focus of computer vision and image processing
techniques is in understanding or extracting specific
features from a single image.
Figure 1 Image Mining Processes [1]
Figure 1.1 shows the image mining process. The
images from an image database are first preprocessed
to improve their quality. These images then undergo
various transformations and feature extraction to
generate the important features from the images.
With the generated features, mining can be carried
out using data mining techniques to discover
significant patterns. The resulting patterns are
evaluated and interpreted to obtain the final
knowledge, which can be applied to applications. [1]
2.0 PROBLEM DEFINITION
In the colour based image retrieval the RGB Colour
model is used. Colour images normally are in three
dimensional. RGB colour components are taken from
each and every image. Then, the mean values of Red,
Green, and Blue components of target images are
calculated and stored in the database. Based on the
RGB component mean values, the images are
clustered as Red, Green and Blue major component
categories. These three mean values for each image
are stored and considered as features.
Then the top ranked images are re-grouped according
to their texture features. In the texture-based
approach the parameters gathered are on the basis of
statistical approach. Statistical features of grey levels
were one of the efficient methods to classify texture.
The Grey Level Co-occurrence Matrix (GLCM) is
used to extract second order statistics from an image.
GLCMs have been used very successfully for texture
calculations [9]. The different texture parameters like
entropy, contrast, dissimilarity, homogeneity,
standard deviation, mean, and variance of both query
image and target images are calculated. From the
calculated values the required image from the
repository is extracted.
Then, the pre-processed images in the database are
classified as low-texture, average-texture and high-
texture detailed images respectively based on some
factor like MLE (Maximum Likelihood Estimation)
estimation. The classified images are then subject to
colour feature extraction. The retrieved result is pre-
clustered by Fuzzy-C means technique. This is
followed by GLCM texture parameter extraction
where the texture factors like contrast, correlation,
mean, variance and standard variance are mined. The
resulted values of both the query image and target
images are compared by Euclidean distance method.
2.1 Proposed Solution
In this, a new method for image classification is
formulated in order to reduce the searching time of
images from the image database. The coarse content
of image is grouped under three categories as:
(i) High-texture detailed
Image
(ii) Average-texture detailed
Image
(iii) Low-texture detailed
Image
Thereby, we can reduce the search space by one third
of what was earlier. If we go more number of groups
or less number of groups, they may reveal
unnecessary overlapping overhead problems or may
produce approximate results.
So, the main focus on this classification is by making
use of “textures” present in an image. This is because
this texture-based classification is simple, easy and
efficient for real time applications as compared to
classifications based on Entropy method as well as
segmentation based techniques.
2.2 Image Retrieval
Image Retrieval from the image collections involved
with the following steps
• Pre-processing
• Image Classification based on some
true factor
• RGB Components processing
• Preclustering
• Texture feature extraction
Pre-
process
the image
contents
Mining the
collected
data
Interpretation
& Evaluation
Images in
the
Database
Create
Knowled
2010 IEEE International Conference on Computational Intelligence and Computing Research
• Similarity comparison
• Target image selection
2.3 IMAGE RETRIEVAL SYSTEM
Figure 2.0 Block Diagram of Image
Retrieval System
2.4 Pre-processing & Noise Reduction Filtering
Pre-processing is the name used for operations on
images at the lowest level of abstraction. The aim of
the pre-processing is an improvement of the image
that suppresses unwilling distortions or enhances
some image features, which is important for future
processing of the images. This step focuses on image
feature processing. Filtering is a technique for
modifying or enhancing an image. The image is
filtered to emphasize certain features or remove other
features. The noise in the images is filtered using
linear and non-linear filtering techniques. Median
filtering is used here to reduce the noise.[12]
Figure 2.1 Results for Pre-processing Image
2.6 RGB Components Processing
An RGB colour images is an M*N*3 array of colour
pixels, where each colour pixel is a triplet
corresponding to the red, green, and blue components
of an image at a spatial location. An RGB image can
be viewed as the stack of three gray scale images
that, when fed into the red, green, blue inputs of a
colour monitor, produce the colour image on the
screen. By convention the three images form an RGB
images are called as red, green and blue components.
The average values for the RGB components are
calculated for all images
Red average= sum of all the Red Pixels in the image R (P)
No. Of pixels in the image P
Green average= sum of all the Green Pixels in the image G (P)
No. Of pixels in the image P
B average= sum of all the Blue Pixels in the image B (P)
No. Of pixels in the image P
Where R (P) = RED component pixels,
G (P) = GREEN component pixels,
B (P) = BLUE component pixels,
P =No. of pixels in the image.
After calculating the mean values of Red, Blue and
Green components, the values are to be compared
with each other in order to find the maximum value
of the components. For eg., if the value of Red
component is High than the rest of the two, then we
can conclude that the respective image is Red
Intensity oriented image and which can be clustered
into Red Group of Images.
Whenever the query image is given,
calculate the RGB components average values. Then
compare this with the stored values.
Figure 2.2 Result of RGB Components Clustering
Images
2.7 Entropy Classification
The texture represents the energy content of the
image. If an image contains more and high textures,
then the energy will be high as compared to that of
average and low texture images. There are several
texture parameters to be considered [12]. However,
here, the texture parameter Entropy is highly focused
Query &
Target Images
Preprocessing &
Classification
RGB Components
Processing
Clustering Based
on RGB
Components
Texture
Calculation
for images
and
Clustering
Sort out the
Results
Entropy
Calculation
Select Target
Image
2010 IEEE International Conference on Computational Intelligence and Computing Research
and which is to be calculated for the query and target
images. Entropy is a statistical measure of
randomness that can be used to characterize the
texture of the input image. Entropy is defined as
-sum(hc.*log2(hc))
where hc is the histogram counts obtained from the
histogram calculation.
Figure 2.3 Processes for High and Average Texture
Analysis
2.8 Image Clustering
Clustering will be more advantage for reducing the
searching time of images in the database. Fuzzy C-
means (FCM) is one of the clustering methods which
allow one piece of data to belong to two or more
clusters. In this clustering, each point has a degree of
belonging to clusters, as in fuzzy logic, rather than
belonging completely too just one cluster. Thus,
points on the edge of a cluster may be in the cluster
to a lesser degree than points in the centre of cluster.
FCM groups data in specific number of clusters.
Figure 2.4 Results of Clustering Processes
2.9 Similarity Comparison and Image Retrieval
The given query image is pre-processed and the
features of the given query image to be calculated in
the usual way. Then, the entropy value of the given
query image is calculated based on the calculation
given in Sec.2.7. A threshold constant value is to be
added with the entropy value of the query image.
Hence, the result will be compared with the
concerned cluster and the target images are to be
retrieved based on the constraints. The results of this
process are shown in the figure 2.5.
Figure 2.5 Results of Image Retrieval for the given
Query Image
3.0 Performance Evaluation of Proposed CBIR
System
Evaluation of retrieval performance is a crucial
problem in Content-Based Image Retrieval (CBIR).
Many different methods for measuring the
performance of a system have been created and used
by researchers. We have used the most common
evaluation methods namely, Precision and Recall
2010 IEEE International Conference on Computational Intelligence and Computing Research
usually presented as a Precision vs Recall graph.
Precision and recall alone contain insufficient
information. We can always make recall value 1 just
by retrieving all images. In a similar way precision
value can be kept in a higher value by retrieving only
few images or precision and recall should either be
used together or the number of images retrieved
should be specified.
With this, the following formulae are used for finding
Precision and Recall values.
No. of Relevant Images Retrieved
Precision=___________________________
Total number of Images Retrieved
databasetheinimagesrelevantofnoTotal
retrievedimagesrelevantofNo
call
.
.
Re =
4.0 CONCLUSION
The main objective of the image mining is to remove
the data loss and extracting the meaningful
information to the human expected needs. The
images are preprocessed with various techniques and
the texture calculation is highly focused. Here,
images are clustered based on RGB Components,
Texture values and Fuzzy C mean algorithm. Entropy
is used to compare the images with some threshold
constraints. This application can be used in future to
classify the medical images in order to diagnose the
right disease verified earlier.
REFERENCES
1. Image Mining: Trends and Developments, Ji
Zhang Wynne Hsu Mong Li Lee
2. U. M. Fayyad, S. G. Djorgovski, and N. Weir:
Automating the Analysis and ataloging of Sky
Surveys. Advances in Knowledge Discovery
and Data Mining, 471-493, 1996.
3. W. Hsu, M. L. Lee and K. G. Goh. Image
Mining in IRIS: Integrated Retinal
Information System (Demo), in Proceedings
of ACM SIGMOD International Conference
on theManagement of Data, Dallas, Texas,
May 2000.
4. A. Kitamoto. Data Mining for Typhoon Image
Collection. In Proceedings of the
SecondInternational Workshop on Multimedia
Data Mining (MDM/KDD'2001), San
Francisco, CA,USA, August, 2001.
5. C. Ordonez and E. Omiecinski. Discovering
Association Rules Based on Image Content.
Proceedings of the IEEE Advances in Digital
Libraries Conference (ADL'99), 1999.
6. O. R. Zaiane, J. W. Han et al. Mining
MultiMedia Data. CASCON'98: Meeting of
Minds, pp 83-96, Toronto, Canada, November
1998.
7. M. C. Burl et al. Mining for Image Content. In
Systemics, Cybernetics, and Informatics /
Information Systems: Analysis and Synthesis,
(Orlando, FL), July 1999.
8. M. Datcu and K. Seidel. Image Information
Mining: Exploration of Image Content in
LargeArchives. IEEE Conference on
Aerospace, Vol.3, 2000.
9. Yixin Chen, James Z.Wang, Robert Krovetz
“Cluster Based Retrieval Of Images by
Unsupervised Learning”, IEEE Transaction on
Image Processing, Vol 14, pp.1187-1199,
No.8, August 2005.
10. D.S. Zhang and G.Lu, “Content Based Image
Retrieval Using Texture Features”, In Proc. of
First IEEE Pacific-Rim Conference on
Multimedia (PCM’00) pp.392-395, Sydney,
Australia, December2000.
11. Hewayda M.Lofty, Adel S.Elmaghraby
“CoIRS: Cluster Oriented Image Retrieval
System” IEEE conference on tools with
Artificial Intelligence, 2004.
12. “Comparison of Feature Selection Techniques
for Detection of Malignant Tumor in Brain
Images”, M. Sasikalal and N. Kumaravel2,
IEEE Indicon 2005 Conference, Chennai,
India, 11 -13 Dc 2005