A Survey On: Content Based Image Retrieval Systems Using Clustering Techniques For Large Data sets

International Journal of Managing Information Technology (IJMIT) Vol.3, No.4, November 2011
DOI : 10.5121/ijmit.2011.3403 23
A Survey On: Content Based Image Retrieval
Systems Using Clustering Techniques For Large
Data sets
Mrs Monika Jain1
, Dr. S.K.Singh2
1Research scholar, Department of computer science, Mewar university, Rajasthan, India.
monika_smec@yahoo.co.in
2Professor and Head of Department of Information Technology, HRIT Engineering
college, Ghaziabad, India.
singhsks123@gmail.com
ABSTRACT
Content-based image retrieval (CBIR) is a new but widely adopted method for finding images from vast
and unannotated image databases. As the network and development of multimedia technologies are
becoming more popular, users are not satisfied with the traditional information retrieval techniques. So
nowadays the content based image retrieval (CBIR) are becoming a source of exact and fast retrieval. In
recent years, a variety of techniques have been developed to improve the performance of CBIR. Data
clustering is an unsupervised method for extraction hidden pattern from huge data sets. With large data
sets, there is possibility of high dimensionality. Having both accuracy and efficiency for high dimensional
data sets with enormous number of samples is a challenging arena. In this paper the clustering techniques
are discussed and analysed. Also, we propose a method HDK that uses more than one clustering technique
to improve the performance of CBIR.This method makes use of hierachical and divide and conquer K-
Means clustering technique with equivalency and compatible relation concepts to improve the performance
of the K-Means for using in high dimensional datasets. It also introduced the feature like color, texture and
shape for accurate and effective retrieval system.
General Terms
Content Based Image Retrieval , divide and conquer k-means, hierarchical
KEYWORDS
Content Based Image Retrieval, Color, Texture, shape
1. INTRODUCTION
Content-Based Image Retrieval (CBIR) is defined as a process that searches and retrieves images
from a large database on the basis of automatically-derived features such as color, texture and
shape. The techniques, tools and algorithms that are used in CBIR, originate from many fields

24
such as statistics, pattern recognition, signal processing, and computer vision. It is a field of
research that is attracting professionals from different industries like crime prevention, medicine,
architecture, fashion and publishing. The volume of digital images produced in these areas has
increased dramatically over the past 10 decades and the World Wide Web plays a vital role in this
upsurge. Several companies are maintaining large image databases, where the requirement is to
have a technique that can search and retrieve images in a manner that is both time efficient and
accurate (Xiaoling, 2009).In order to meet these requirements, all the solutions, in general,
perform the retrieval process in two steps. The first step is the ‘feature extraction’ step, which
identifies unique signatures, termed as feature vector, for every image based on its pixel values.
The feature vector has the characteristics that describe the contents of an image. Visual features
such as color, texture and shape are more commonly used in this step. The classification step
matches the features extracted from a query image with the features of the database images and
groups images according to their similarity. Out of the two steps, the extraction of features is
considered most critical because the particular features made available for discrimination directly
influence the efficacy of the classification task (Choras, 2007).
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. The focus is to build a universal CBIR system using low level features. Users can query
example images based on these features. 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.
The retrieval of content based image involves the following systems
A. Color-based retrieval
Out of the many feature extraction techniques, color is considered as the most dominant and
distinguishing visual feature. Generally, it adopt histograms to describe it. A color histogram
describes the global color distribution in an image and is more frequently used technique for
content-based image retrieval (Wang and Qin, 2009) because of its efficiency and effectiveness.
Color histograms method has the advantages of speediness, low demand of memory space and
not sensitive with the image’s change of the size and rotation, it wins extensive attention
consequently.
B. The retrieval based on texture feature
The identification of specific textures in an image is achieved primarily by modeling texture as a
two-dimensional gray level variation. Textures are characterized by differences in brightness with
high frequencies in the image spectrum. They are useful in distinguishing between areas of
images with similar color (such as sky and sea, or water, grass). A variety of methods has been
used for measuring texture similarity; the best- established depend on comparing values of what
are well-known as second-order statistics estimated from query and stored images. Essentially,
these estimate the relative brightness of picked pairs of pixels from each image. From these it is

25
possible to measures the image texture such as contrast, coarseness, directionality and regularity
[3] or periodicity, directionality and randomness [4].
C. The retrieval based on shape feature
Shape information are extracted using histogram of edge detection. Techniques for shape feature
extraction are elementary descriptor, Fourier descriptor, template matching, Quantized
descriptors, Canny edge detection [5] etc. Shape features are less developed than their color and
texture counterparts because of the inherent complexity of representing shapes. In particular,
image regions occupied by an object have to be found in order to describe its shape, and a number
of known segmentation techniques combine the detection of low-level color and texture features
with region-growing or split-and-merge processes. But generally it is hardly possible to precisely
segment an image into meaningful regions using low-level features due to the variety of possible
projections of a 3D object into 2D shapes, the complexity of each individual object shape, the
presence of shadows, occlusions, non-uniform illumination, varying surface reflectivity, and so
on.[6]
D. The retrieval based on clustering techniques
Clustering techniques can be classified into supervised (including semi-supervised) and
unsupervised schemes. The former consists of hierarchical approaches that demand human
interaction to generate splitting criteria for clustering. In unsupervised classification, called
clustering or exploratory data analysis, no labeled data are available [9],[10]. The goal of
clustering is to separate a finite unlabeled data set into a finite and discrete set of “natural,”
hidden data structures, rather than provide an accurate characterization of unobserved samples
generated from the same probability distribution [11], [12]. This paper critically reviews and
summarizes different clustering techniques.
D.1. Relevance feedback:
A relevance feedback approach allows a user to interact with the retrieval algorithm by providing
the information of which images user thinks are relevant to the query [13][14][16].Keyword
based image retrieval is performed by matching keyword according to user input and the images
in the database. Some images may not have appropriate keywords to describe them and therefore
the image search will become complex. One of the solution in order to overcome this problem is
“relevance feedback” technique [17] that utilize user feedback and hence reduces possible errors
and redundancy [18][19] .This technique uses a Bayesian classifier [20][13] which deals with
positive and negative feedback. Content based clustering methods cannot adopt to user changes,
addition of new topics due to its static nature. To improve the performance of information
retrieval log-based clustering approaches are brought into the application.
D.2. Log –Based Clustering:
Images can be clustered based on the retrieval system logs maintained by an information retrieval
process [21]. The session keys are created and accessed for retrieval. Through this the session
clusters are created. Each session cluster generates log –based document and similarity of image
couple is retrieved. Log –based vector is created for each session vector based on the log-based
documents[22]. Now, the session cluster is replaced with this vector. The unaccessed documents
creates its own vector.

26
A hybrid matrix is generated with at least one individual document vector and one log-based
clustered vector. At last the hybrid matrix is clustered. This technique is difficult to perform in the
case of multidimensional images. To overcome this hierarchical clustering is adopted.
D.3 Hierarchical Clustering
Hierarchical clustering (HC) algorithms organize data into a hierarchical structure according to
the proximity matrix. The results of HC are usually depicted by a binary tree or dendrogram as
shown in Fig 1where A, B, C, D, E, F, G are objects or clusters. It represents the nested grouping
of patterns and similarity levels at which groupings change. The root node of the dendrogram
represents the whole data set and each leaf node is regarded as a data object. The intermediate
nodes, thus, describe the extent that the objects are proximal to each other; and the height of the
dendrogram usually expresses the distance between each pair of objects or clusters, or an object
and a cluster. The ultimate clustering results can be obtained by cutting the dendrogram at
Fig 1. The dendogram obtained using HC algorithm.
different levels Ref [64]. This representation provides very informative descriptions and
visualization for the potential data clustering structures, especially when real hierarchical
relations exist in the data, like the data from evolutionary research on different species of
organizms. HC algorithms are mainly classified as agglomerative methods and divisive methods.
Agglomerative clustering starts with clusters and each of them includes exactly one object. A
series of merge operations are then followed out that finally lead all objects to the same group.
Divisive clustering proceeds in an opposite way. In the beginning, the entire data set belongs to a
cluster and a procedure successively divides it until all clusters are singleton clusters. For a
cluster with objects, there are 2N-1
-1 possible two-subset divisions, which is very expensive in
computation [30]. Therefore, divisive clustering is not commonly used in practice. In recent
years, with the requirement for handling large-scale data sets in data mining and other fields,
many new HC techniques have appeared and greatly improved the clustering performance.
Typical examples include CURE [65], ROCK [66],Chameleon [67], and BIRCH [68].

27
D.4. Retrieval Dictionary Based Clustering
A rough classification retrieval system is formed. This is formed by calculating the distance
between two learned patterns and these learned patterns are classified into different clusters
followed by a retrieval stage. The main drawback addressed in this system is the determination of
the distance.
To overcome this problem a retrieval system is developed by retrieval dictionary based clustering
[23]. This method has a retrieval dictionary generation unit that classifies learned patterns into
plural clusters and creates a retrieval dictionary using the clusters. Here, the image is retrieved
based on the distance between two spheres with different radii. Each radius is a similarity
measure between central cluster and an input image. An image which is similar to the query
image will be retrieved using retrieval dictionary.
D.5. NCut Algorithm
Ncut method attempts to organize nodes into groups so that the within the group similarity is
high, and/or between the groups similarity is low. This method is empirically shown to be
relatively robust in image segmentation [24]. This method can be recursively applied to get more
than two clusters. In this method each time the subgraph with maximum number of nodes is
partitioned (random selection for tie breaking). The process terminates when the bound on the
number of clusters is reached or the Ncut value exceeds some threshold T. The recursive Ncut
partition is essentially a hierarchical divisive clustering process that produces a tree[25].
Nonetheless, the tree organization here may misleading a user because there is no guarantee of
any correspondence between the tree and the semantic structure of images. Furthermore,
organizing image clusters into a tree structure will significantly complicate the user interface.
D.6. K Means clustering
This nonhierairchal method initially takes the number of components of the population equal to
the final required number of clusters. In this step itself the final required number of clusters is
chosen such that the points are mutually farthest apart. Next, it examines each component in the
population and assigns it to one of the clusters depending on the minimum distance. The
centroid's position is recalculated everytime a component is added to the cluster and this
continues until all the components are grouped into the final required number of clusters.The K-
means algorithm is very simple and can be easily implemented in solving many practical
problems. It can work very well for compact and hyperspherical clusters. The time complexity of
K-means is O(NKd). Since K and d are usually much less than N,K-means can be used to cluster
large data sets. Parallel techniques for K-means are developed that can largely accelerate the
algorithm [70], [71], [72]. Incremental clustering techniques for example (Bradley et al., 1998) do
not require the storage of the entire data set, and can handle it in a one-pattern-at-a-time way. If
the pattern displays enough closeness to a cluster according to some predefined criteria, it is
assigned to the cluster. Otherwise, a new cluster is created to represent the object.
D.7 Graph theory based clustering
The concepts and properties of graph theory [73] make it very convenient to describe clustering
problems by means of graphs. Nodes of a weighted graph correspond to data points in the pattern
space and edges reflect the proximities between each pair of data points. A graph-based clustering
method is particularly well suited for dealing with data that is used in the construction of
minimum spanning tree MST. It can be used for detecting clusters of any size and shape without
specifying the actual number of clusters. Well known algorithms in clustering are Zhan’s

28
Minimum Spanning Tree based clustering [74], [75], and clustering editing method [78] [79],
HCS algorithm [80], etc. Current research is focused on clustering using divide and conquers
approach [81]. Usually this clustering methodology is used to detect irregular clustering
boundaries in clustering results. Zhan [74] proposes to construct an MST and delete the
inconsistent edges, i.e. the edges weight values are significantly larger than average weight of the
nearby edges in the tree. The inconsistency measure [82] is applied to each edge to detect and
remove the inconsistence edges, which results as a set of disjoint subtrees, each subtree will
represent a separate cluster
D.8. Divide and Conquer K-Means
When the size of a data set is too large, it is possible to divide the data into different subsets and
to use the selected cluster algorithm separately to these subsets. This approach is known as divide
and conquer [26, 27]. The divide and conquer algorithm first divides the entire data set into a
subset based on some criteria. The selected subset is again clustered with a clustering algorithm
K-Means. The advantage is to accelerate search and to reduce complexity which depends on
number of samples. Methods based on subspace clustering may help to ease the problem of
clustering high-dimensional data, but they are not adapted at obtaining a large number of clusters
[28][29].A possible solution to this issue, is to cluster hierarchically (obtain a small number of
clusters and then cluster again each of the clusters obtained). The proposed enhanced clustering
method HDK which uses the combination of unsupervised clustering methods is one of the
method that can largely accelerate the CBIR system. In next section problem definition and
framework describes. In sections 3,4,5 and 6 evolution , recent work ,open areas and then
conclusion are given respectively.
2. PROBLEM DEFINITION
Clustering algorithms partition data into a certain number of clusters (groups, subsets, or
categories). There is no universally agreed upon definition [30]. Most researchers describe a
cluster by considering the internal homogeneity and the external separation [31], [32], [33], i.e.,
patterns in the same cluster should be similar to each other, while patterns in different clusters
should not. Both the similarity and the dissimilarity should be examinable in a clear and
meaningful way. Here, we give some simple mathematical descriptions of several types of
clustering, based on the descriptions in [32]. Given a set of input patterns X={x1,….,xj,….,xN},
where xj=(fj1,fj2,….fjd)T
Є Rd
and each measure fji is said to be a feature (attribute, dimension, or
variable).
• (Hard) partitional clustering attempts to seek a K–partition of X,C={C1,….,CK}(K≤ N), such
that
1) Ci ≠ φ , i=1,…,K;
2) K
i 1
=
U Ci = X;
3) Ci I Cj = φ ,I,j=1,….,K and i ≠ j .

29
2.1 Image Retrieval
Image Retrieval using HDK algorithm from the image collections involved with the following
steps.
1. Pre-processing is based on RGB color Components using Hierarchical clustering Method.
2. Apply divide and conquer k means.
Some hypothesis has been considered:
• H1 - Proposed method would be able to group samples of same no of clusters and find
similarity among them.
• H2 - Proposed method is faster and accurate than single step clustering due to use of
divide and conquers technique.
• H3 – Proposed method would allow Euclidean distance to be used in high dimensional
data.
In this study we assume that the space is orthogonal and dimensions for all objects are the same
and finally we use ordinal data type because of the application. We group images based on
number of clusters and clusters are retrieved based on color which is one of the most widely used
features for image similarity retrieval. The advantage is nearly linear trend means stability in
execute of time . The proposed algorithm is scalable because it follows K-Means complexity that
is linear with n number of samples, T number of iteration, d number of dimensions and k number
of clusters: O(T.K.N.D). Both HC and K-Means use pair’s similarity measurement that is very
time consuming. By dividing space into subspaces, run time would be reduced because pair’s
similarity measurement doesn’t need to apply for whole samples.
2.2 Framework
In this study, our framework includes three main steps as shown in fig.2.
1. Preprocessing is done by applying hierarchical clustering algorithm based on
color feature.
2. After preprocessing find the no of clusters for subspace sampling
3. And in final step we separately cluster each subspaced sampled using k means
clustering technique.
Fig.2 Framework for clustering high dimensional data set.
1. Preprocessing
2.Divide subspace
clustering
3.Cluster the subspace
clustering

30
3.EVOLUTION
In the early 1990s, as a result of advances in the Internet and new digital image sensor
technologies, the volume of digital images produced by scientific, educational, medical,
industrial, and other applications available to users increased dramatically. Most early work on
CBIR was based on computing color, shape, and texture based features and using them to define a
similarity between the images. A 2008 survey on CBIR highlights the different approaches used
for CBIR through time (Datta et al., 2008). Recent approaches for CBIR use key point based
features .For example, SIFT (Lowe, 2004) descriptors can be used to represent the images.
However, once the size of the image database increases (10 million), and assuming 10 ms to
compute the matching score between an image pair, a linear search would take approximately 30
h to answer one query. This clearly is unacceptable. A novel approach for image retrieval is to
convert the problem into a text retrieval problem. The key points from all the images are first
clustered into a large number of clusters (which is usually much less than the number of key
points themselves). These are called visual words. An image is then represented by a histogram of
visual words, i.e., the number of keypoints from the image that are in each word or each cluster.
The difficulties faced by text based retrieval became more and more severe. The efficient
management of the rapidly expanding visual information became an urgent problem. Greg Pass
Ramin Zabih [34] in 1996,described histogram refinement for comparing images . Histogram
refinement splits the pixels in a given bucket into several classes, based upon some local
property.Within a given bucket, only pixels in the same class are compared. Here describe a split
histogram called a color coherence vector (CCV), which partitions each histogram bucket based
on spatial coherence. After that in 1997, Chad Carson, Serge Belongie, Hay it Greenspan, and
Jitendra Malik [35] ,they present a new image representation which provides a transformation
from the raw pixel data to a small set of localized coherent regions in color and texture space.
This so-called “blobworld” representation is based on segmentation using the Expectation
Maximization (EM) algorithm on combined color and texture features. Minakshi Banerjee, Malay
K.Kundu [36] in 2003 discussed the common problem in content based image retrieval (CBIR) is
selection of features. Image characterization with lesser number of features involving lower
computational cost is always desirable. Color is an important attribute for image matching and
retrieval. Recently there has been an increased interest in color research, e.g. classification and
segmentation. Swain and Ballard [37] have proposed a color matching method in their paper on
'color indexing'. Their method, known as histogram intersection, is based on matching of color
histograms and the core idea in their technique is to compute:
H(I,M)=
∑
∑
=
=
n
j
j
n
j
j
j
M
M
I
1
1
)
,
min(
where H(I,M) is the match value and I and M are image (query image) and model (an image in
the database) histograms, respectively, each containing n bins. The match value is computed for
every model histogram and the value is closer to unity if the model image is similar to the query
image.
In 1995, Ref[38] Mohan S. Kankanhalli,T Babu M. Mehtre And Jian Kang Wut proposed new
method in which color clustering can be used to find out clusters and to assign a representative

31
color to each of these clusters. For each pixel, compute the color distance to the different clusters.
Assign the pixel to the cluster for which color distance is minimum. Thus, every pixel is assigned
to one of the clusters. Unfortunately, for a given set of color images, there is not enough
information about the number and population of clusters.
Ref[28] feature selection methods select only the most relevant of the dimensions from a dataset
to reveal groups of objects that are similar on only a subset of their attributes. While quite
successful on many datasets, feature selection algorithms have difficulty when clusters are found
in different subspaces. It is this type of data that motivated the evolution to subspace clustering
algorithms. These algorithms take the concepts of feature selection one step further by selecting
relevant subspaces for each cluster separately. All clustering algorithms require that the user set
some parameters and make some assumptions about the clusters to be discovered. For detecting
data subsets combined with a set of dimensions suitable for local reduction, subspace clustering is
used. Clustering in general is used to find natural object groupings in data. The objects contained
in such groups are good candidates for generating minimal bounding regions in index structures.
Subspace clustering goes beyond fullspace clustering: It was proposed first in [agarwal 1998]. A
subspace cluster is a grouping of objects that are only similar in a subset of their dimensions and
thus minimum bounding regions that are induced by subspace clusters are very compact. The
CLIQUE algorithm [93] was one of the first subspace clustering algorithms. The key idea of the
CLIQUE algorithm is that if there are dense units in k dimensions, there are dense units in all (k-
1) dimensional projections ,various subspace clustering algorithms have been published. Some of
them extend CLIQUE in different directions[cheng1999,nagesh1999,woo2002].
Subspace clustering algorithms allow users to break the assumption that all of the clusters in a
dataset are found in the same set of dimensions.
There are many potential applications with high dimensional data where subspace clustering
approaches could help to uncover patterns missed by current clustering approaches. Applications
in bioinformatics and text mining are particularly relevant and present unique challenges to
subspace clustering.
Segmentation is very important to Image Retrieval, shape feature depend on good segmentation.
After extracting features next step in content based image retrieval system is efficient multi-
dimensional techniques. But there are two main difficulties occur in exploring multi-dimensional
indexing techniques for image retrieval system and they are:
1) High dimensionality (The dimensionality of the feature vectors is normally of the
order of 10
2
.
2) Non-Euclidean similarity measure (Since Euclidean measure may not effectively simulate
human perception of a certain visual content, various other similarity measures, such as
Histogram Intersection, Cosine, Correlation, etc. need to be supported).
To overcome these problems, we need to perform dimensional reduction and then use appropriate
multi-dimensional indexing techniques which are capable of supporting Non-Euclidean similarity
measures. The dimension of the feature vectors in image retrieval system is very high, the
embedded dimension is much lower [28] before we utilize any indexing technique, it is beneficial
to first perform dimension reduction. There are two approaches for dimensional reduction
Karhunen-Loeve Transform (KLT) and column-wise clustering . After identifying the embedded

32
dimension of the feature vectors, we need to select appropriate multi-dimensional indexing
algorithms to index the reduced but still high dimensional feature vectors. There are many
popular multi-dimensional indexing techniques present; they are Bucketing algorithm-d tree,
priority k-d tree [39], quard-tree, K-D-B tree, hb-tree, R-tree and its variants R+
-tree and R*
-tree
[83, 84, 85, 86, 87]. In addition to the above approaches, clustering and neural nets, widely used
in pattern recognition, are also promising indexing techniques [88, 89]. With highdimensional
data, indexing schemes such as k-d trees do not work well, to the point where examining every
data point (i.e. using no special data structure for acceleration) can be much faster than algorithms
intended for low-dimensional acceleration. The most popular and the simplest partitional
algorithm is K-means. K-means has a rich and diverse history as it was independently discovered
in different scientific fields by Steinhaus (1956), Lloyd (proposed in 1957, published in 1982),
Ball and Hall (1965), and MacQueen (1967). Even though K-means was first proposed over 50
years ago, it is still one of the most widely used algorithms for clustering. Ease of
implementation, simplicity, efficiency, and empirical success are the main reasons for its
popularity. In 1988, Jain and Dubes gives simple steps of k-means algorithm. Since partitional
algorithms are preferred in pattern recognition due to the nature of available data. Although the
topic has been studied for the past 15 years or so, there has been only limited success. The most
efficient current method for high-dimensional k-means clustering is Elkan’s algorithm [97]. His
method does not use any indexing structure, but keeps a number of distance bounds that allow it
to avoid unnecessary distance computations. Greg Hamerly proposed another modified and
simplified of Elkan’s k-means algorithm. As such, he used efficiently updated distance bounds
and the triangle inequality to avoid point-center distance calculations.
Because most clustering applications tend to apply to a massive size of datasets, we are interested
in finding a method to partition a data set by K-means algorithm in such a way that we can
temporarily ignore some resulting partitions in order to work on a few chosen specific partitions
without sacrificing clustering quality. A motivation is to save some significant working space and
we believe, by means of doing so, we can find a way to add a new portion of a data set into the
existing partitions without having to rerun the K-means algorithm from the start. Some partially-
related work has already been studied in [97]. Example algorithms include BIRCH (Zhang et al.,
1996), divide-and-conquer (Steinbach et al., 2000), coreset K-means (Har-peled and Mazumdar,
2004), and coarsening methods (Karypis and Kumar, 1995). Bradley et al. (1998) presented a fast
scalable and singlepass version of K-means that does not require all the data to be fit in the
memory at the same time. X-means (Pelleg and Moore, 2000) automatically finds K by
optimizing a criterion such as Akaike Information Criterion (AIC) or Bayesian information
Criterion (BIC). In K-medoid (Kaufman and Rousseeuw, 2005), clusters are represented using the
median of the data instead of the mean. Kernel K-means (Scholkopf et al., 1998) was proposed to
detect arbitrary shaped clusters, with an appropriate choice of the kernel similarity function. In
2009, Madjid Khalilian, Farsad Zamani Boroujeni,proposed a method k-means divide and
conquer in which objects are clustered base on their size, in fact, they are using subspaces for
clustering Ref[40].
4. RECENT WORK
In a small database ,a simple sequential scan is usually employed for k nearest – neighbor (K-
NN) search. But for large data set , efficient indexing algorithms are imperative. High
dimensional data is increasingly in many common fields. As the number of dimensions increase,
many clustering techniques begin to suffer from the curse of dimensionality, degrading the

33
quality of the results. In high dimensions, data becomes very sparse and distance measures
become increasingly meaningless. There is a general categorization for high dimensional data set
clustering:1-Dimension reduction, 2-Parismonious models, 3-Subspace clustering [44]. Feature
selection and feature extraction are most popular techniques in dimension reduction. It is clear
that in both methods we will have losing information which naturally affects accuracy. Ref. [45]
reviewed the literature on parsimonious models and Gaussian models from the most complex to
simplest which yields a method similar to the K Means approach. When we have low dimensional
spaces these methods aren’t able to work well. There are two main approaches for subspaces
methods: in first class centers are considered on a same unknown subspace and in second each
class is located on specific subspace [46]. The idea of subtopics or subgroups is appropriate for
document clustering and text mining [47]. Tensor factorization as a powerful technique has been
used in [48]. Inconsistency has been shown in those public data sets because of outlier. Ref. [49]
proposed a framework which integrates subspace selection and clustering. Equivalency between
kernel K-Means clustering and iterative subspace selection has been shown. Ref. [50] proposed a
general clustering improver scheme which is included two main steps. In first step it uses intrinsic
properties of the data set for dimension reduction after that several iteration of a clustering
algorithms are applied, each with different parameter. Base on BIC criterion the best result will
be selected. There are some weaknesses for this method e.g. since BIC fits a model to specific
data distribution it cannot be used to compare models of different data sets. Ref. [51] presented a
semi-supervised clustering method base on spherical K-Means via feature projection which is
tailored for handling sparse high dimensional data. They first formulated constraint-guided
feature projection then applied the constraint spherical K-Means algorithm to cluster data with
reduced dimension. Ref.[52] proposed two methods of combining objective function clustering
and graph theory clustering.– Method 1: incorporates multiple criteria into an objective function
according to their importance, and solves this problem with constrained nonlinear optimization
programming.– Method 2: consists of two sequential procedures:
(a) A traditional objective function Clustering for generating the initial result
(b) An auto associative additive system based on graph theory clustering for modifying the initial
result.
Ref. [53] Proposed sequential combination methods for data clustering
– In improving clustering performance they proposed the use of more than one clustering
method.
– They investigated the use of sequential combination clustering as opposed to simultaneous
combination and found that sequential combination is less complex and there are improvements
without the overhead cost of simultaneous clustering. In clustering points lying in high
dimensional spaces, formulating a desirable measure of “similarity” is more problematic. Recent
research shows that for high dimensional spaces computing the distance by looking at all the
dimensions is often useless, as the farthest neighbor of a point is expected to be almost as close as
its nearest neighbor [54]. To compute clusters in different lower dimensional subspaces, recent
work has focused on projective clustering, defined as follows: given a set P of points in Rd and an
integer K, partition into subsets that best classify into lower dimensional subspaces according to
some objective function. Instead of projecting all the points in the same subspace, this allows
each cluster to have a different subspace associated with it [55]. Proposed model will be more
effective and achieves significant performance improvement over traditional method for most

34
clustering. It will be able to cluster samples of same no of clusters and level and be more efficient
and accurate than a single one pass clustering. There are some delimitations for this method. First
space should be orthogonal it means there is no correlation among attributes of an object. Second
base on application that is used all attributes in an object have the same kind of data types.
Without losing generality it is possible to extend proposed method to other kinds of data types.
Samples are considered in a same number of dimensions. The proposed method HDK select
subspaces and perform clustering based on these subspaces.
Fred and Jain [56] use a hierarchical clustering with a single-link method is then applied to the
co-occurrence matrix. The final number of clusters is taken as the one that corresponds to the
longest lifetime on the dendrogram of the hierarchical algorithm. The method of Fred and Jain is
well adapted when the number of clusters is approximately known a priori. If the number of
clusters is sequentially changed from 2 to the number of samples, the co-association matrix will
change towards a near diagonal matrix with small values out of diagonal. So, the more clusters
used to build the coassociation matrix the more clusters result from the combination.The same
kind of experiments but with another combination scheme is made by Topchy et al. in [58].Lange
and Buhmann [59] optimize a probabilistic model of the co-association matrix. The EMalgorithm
optimizes model parameters and needs O(I2) operations for each iteration, where I is the number
of data samples. It makes difficult to apply this approach to a high volume of data. Many of the
presented methods need to know a priori about data to combine clusterings or to set manually
some parameters for a combination scheme. This provided us with the motivation to pose the
problem of combination in a form which will not depend on any parameter and a priori
knowledge. Our formulation of the problem is based on a co-association matrix. It allows us to
process a huge volume of data as well as clusterings.
4. OPEN AREAS
Similarity search is an important topic of research, which finds many applications and not
surprisingly there is a vast literature about the subject. It is already been discussed that the
existing techniques may be used to improve the quality of image retrieval and the understanding
of user intentions. An approach that combines two different approaches to image retrieval,
together with active use of context information and interaction has been proposed. An important
aspect of the research work outlined here is to design and evaluate a system that, in addition to
combine the use of TBIR with CBIR [60]. Use of the hybrid feature including color, texture and
shape as feature vector of the regions to match images can give better results. Results on a
database of 1000 general-purposed images demonstrate the efficiency and effectiveness of the
image representation for region based image retrieval [61]. The technique called self taught
multiple-instance learning (STMIL) [62] that deals with learning from a limited number of
ambiguously labeled examples is also a effective area to work with for efficient results. The
problem of bridging the semantic gap between high level query which is normally in terms of an
example image and low level features of an image such as colour, texture, shape and object
forced to apply techniques to reduce the semantic gap. 3D informatics[90] incorporates additional
research areas including content-based retrieval, image understanding, indexing, data mining, and
data management. Existing techniques for image retrieval based on divide and conquer k-means
and subspace clustering is a novel approach for retrieving the images based on their contents.
HDK algorithm can be used in various areas to improve the performance of the system and to
achieve better results in different applications.

35
5. CONCLUSIONS
The purpose of this survey is to provide an overview of the functionality of content based image
retrieval systems. Combining advantages of HC and divide and conquer K-Means strategy can
help us in both efficiency and quality. HC algorithm can construct structured clusters. Although
HC yields high quality clusters but its complexity is quadratic and is not suitable for huge datasets
and high dimension data. In contrast K-Means is linear with size of data set and dimension and
can be used for big datasets that yields low quality. Divide and conquer K-Means can be used for
high dimensional data set . In this paper we present a method HDK to use both advantages of HC
and Divide and conquer K-Means by introducing equivalency and compatible relation concepts.
Using two steps clustering in high dimensional data sets with considering no of clusters based on
color feature helps us to improve accuracy and efficiency of original K-Means clustering. For this
purpose we should consider orthogonal space. HDK algorithm has been used extensively in
various areas to improve the performance of the system and to achieve better results in different
applications.
REFERENCES:
[1] Rui Xu, Donald Wunsch II, “Survey of Clustering Algorithms”, IEEE Transactions On Neural
Networks, Vol. 16, No. 3, May 2005.
[2] Ritendra Datta, Dhiraj Joshi, Jia Li, And James Z. Wang, “Image Retrieval: Ideas, Influences, and
Trends of the New Age”, ACM Computing Surveys, Vol. 40, No. 2, Article 5, Publication date: April
2008.
[3] Tamura et al. “Texture Feautres Corresponding to Visual Perception”- IEEE Trans on system, Man
and cyber 8-460-472-1978.
[4] Sanjoy Kumar Saha et al. “CBIR Using Perception Based Texture And Colour measures ”CSE
Department; CST Department Jadavpur Univ., India; B.E. College, Unit ISI, Kolkata, India - -2003.
[5] Canny, J., “A computational approach to edge detection”, IEEE Trans on Pattern Analysis and
Machine Intelligence, 8:679-698, 1986.
[6] S.Nandagopalan, Dr. B.S. Adiga, and N. Deepak “A Universal Model for Content-Based Image
Retrieval” World Academy of Science, Engineering and Technology 46 2008.
[7] Mohammed Eisa and Ibrahim Elhenawy and A. E. Elalfi and Hans Burkhardt, “Image Retrieval
based on Invariant Features and Histogram Refinement”, ICGST International Journal on Graphics,
Vision and Image Processing, March 2006, pp. 7-11.
[8] V. Castelli and L. D. Bergman (Eds.), “Image Databases: Search and Retrieval of digital Imagery”,
Wiley: New York, 2002.
[9 ] B. Everitt, S. Landau, and M. Leese, “Cluster Analysis”, London:Arnold, 2001.
[10] A. Jain and R. Dubes, “Algorithms for Clustering Data”, Englewood Cliffs, NJ: Prentice-Hall, 1988.
[11] A. Baraldi and E. Alpaydin, “Constructive feedforward ART clustering networks—Part I and II,”
IEEE Trans. Neural Netw., vol. 13, no. 3, pp.645–677, May 2002.
[12] V. Cherkassky and F. Mulier, “Learning From Data: Concepts, Theory, and Methods”, New York:
Wiley, 1998.
[13] I. J. Cox, M. L. Miller, T. P. Minka, T. V. Papathomas and P.Yianilos, “The Bayesian Image
Retrieval System, PicHunter: Theory, Implementation and Psychophysical Experiments”, IEEE
Trans. Image Processing, vol. 9, no. 1, pp. 20-37, 2000.
[14] Y. Rui, T. S. Huang, M. Ortega and S. Mehrotra, “Relevance Feedback: A Power Tool for Interactive
Content-Based Image Retrieval”, IEEE Trans. Circuits and Video Technology, vol. 8, no. 5, pp. 644-
655, 1998.
[15] S. Thilagamani and N. Shanthi , “A Survey on Image Segmentation Through Clustering”,
International Journal of Research and Reviews in Information Sciences , Vol. 1, No. 1, March 2011 .

36
[16] Rui Y, Thomas, S. Huang. “Content-based image retrieval with relevance feedback in MARS”, In
proceedings of IEEE international conference on image processing, pp. 815-818, 1997.
[17] Hoi, C.-H. and Lyu, M. R. 2004b. “A novel log based relevance feedback technique in content based
image retrieval”. In Proc. ACM Multimedia,2004.
[18] Zhou XS, Huang TS., Relevance feedback in image retrieval: A comprehensive review. Multimedia
Syst; 8: 536-544, 2003.
[19] P.Chundi, U.Dayal, , M.Sayal,, M.Hsu, “A Document Clustering Method and System:” google
patents,US20077181678, 2007.
[20] D. Melas and S. Wilson, “Double Markov Random Fields and Bayesian Image Segmentation,” IEEE
Trans. Signal Processing,vol. 50, no. 2, pp. 357-365, Feb. 2002.
[21] Huiyu Zhou, Abdul H. Sadka, Mohammad R. Swash, Jawid Azizi and Abubakar S. Umar., “Content
Based Image Retrieval and clustering: A Brief Survey” school of Engineering and Design, Brunel
University, Uxbridge, UB8 3PH, UK
[22] Hoi, C.-H. and Lyu, M. R. 2004a. “Group-based relevance feeedbacks with support vector machine
ensembles” In Proc. IEEE ICPR,2004.
[23] Li Wenchao Zhou Yong Xia Shixiong China Univ. of Min. & Technol., Xuzhou, “A Novel Clustering
Algorithm Based on Hierarchical and K-means Clustering” Print ISBN: 978-7-81124-055-9, On
page(s): 605, 2009.
[24] Jianbo Shi Malik, J.Robotics Inst., Carnegie Mellon Univ., Pittsburgh, PA , “Normalized cuts and
image segmentation”, ISSN: 0162-8828, On page(s): 888 – 905,2000.
[25] Emma Regentova, Dongsheng Yao, and Shahram Latifi, “Image segmentation using Ncut in the
wavelet domain”, International journal of image and graphics,world scientific Publishing
company,vol.6(4),pp.569-582,2006.
[26] Guha, Meyerson, A. Mishra, N. Motwani, and O. C. ."Clustering data streams: Theory and practice ."
IEEE Transactions on Knowledge and Data Engineering, vol. 15,pp. 515-528, 2003.
[27] A. Jain , M. Murty , and p. Flynn " Data clustering: A review.," ACM Computing Surveys, vol. 31,
pp. 264-323,1999.
[28] L. Parsons, E. Haque and H. Liu., “Subspace clustering for high dimensional data: a review.”
SIGKDD Explor. Newslett. 6 (1), pp. 90–105. 2004.
[29] C. Bouveyron, S. Girard and C. Schmid, “High-dimensional data clustering. Computational Statistics
& Data Analysis”, vol. 52, pp.502–519. 2007.
[30] B. Everitt, S. Landau, and M. Leese, “Cluster Analysis” . London:Arnold, 2001.,
[31] Classification, 2nd ed. London, U.K.: Chapman & Hall, 1999.
[32] P. Hansen and B. Jaumard, “Cluster analysis and mathematical programming”, Math. Program., vol.
79, pp. 191–215, 1997.
[33] A. Jain and R. Dubes, “Algorithms for Clustering Data”. Englewood Cliffs, NJ: Prentice-Hall, 1988.
[34] Greg Pass, Ramin Zabih, “Histogram refinement for content based image retrieval” WACV '96.
[35] Chad Carson, Serge Belongie, Hay it Greenspan, and Jitendra Malik, “Region Based Image
Querying,” this work is supported by an NSF digital library grant (IRI 94-11334) 1997 IEEE.
[36] Minakshi Banerjee 1, Malay K. Kundu “Edge based features for content based image retrieval”
0031-3203/2003.
[37] M. J. Swain and D. H. Ballard, “Color indexing” , Int. d. Comput. Vis. 7(1), 11-32 (1991).
[38] Mohan S. Kankanhalli,T Babu M. Mehtre and Jian Kang Wut, “Cluster-Based Color Matching”,
Pattern Recognition. Vol. 29, No. 4, pp. 701 708, 1996.
[39] Madjid Khalilian, Norwati Mustapha, MD Nasir Suliman, MD Ali Mamat,” A Novel K-Means Based
Clustering Algorithm for High Dimensional Data Sets”,Proceedings of the IMECS 2010.
[40] Madjid Khalilian, Farsad Zamani Boroujeni, Norwati Mustapha, Md. Nasir Sulaiman, “K-Means
Divide and Conquer Clustering”, International Conference on Computer and Automation
Engineering.
[41] H. T. Shen, X. Zhou, and A. Zhou. “An adaptive and dynamic dimensionality reduction method for
high-dimensional indexing” VLDB J., 16(2):219{234,2007.
[42] Y. Tao, K. Yi, C. Sheng, and P. Kalnis., “Quality and efficiency in high dimensional nearest neighbor
search”,.In SIGMOD, pages 563{576, 2009.

37
[43] R. Weber, H.-J. Schek, and S. Blott., “A quantitative analysis and performance study for similarity-
search methods in high-dimensional spaces.”, In VLDB, pages 194{205, 1998.
[44] C.Bouveyran, S. Girard , C.Schmid : Technical Report 1083M, LMC-IMAGE 2008.
[45] C.Fraley and A. Raftery, “Model based clustering discriminate analysis and density estimation”,
journal of American statistics association, 97, 611-631 ,2002.
[46] H. Bock, on the interface between cluster analysis, principal component clustering and
multidimensional scaling, 1987.
[47] Jeffery L. Solka: Text data mining Theory and methods, vol 2, 94-112, 2008
[48] H. Huang, C. Ding, D. Luo, T. Li, “simultaneous tensor subspace selection and clustering”, KDD08,
lasvegas USA ,2008.
[49] J. Ye, Z. Zhao, M. Wu, “Discriminative K-Means for clustering”, proceeding of the annual
conference, 2007.
[50] R. Varshavsky, D. itorn and M. Linial, “Cluster algorithm optimizer: A framework for large datasets”
,ISBRA pp 85-96,springer 2007.
[51] W. Tang, H. Xiong, S. Zhang, J. Wu, “Enhancing semisupervised clustering”, KDD07 california USA
, ACM 2007.
[52] Yuntao Qian and Ching Y. Suen , “Clustering Combination Method”, IEEE 2000.
[53] Yuntao Qian and Ching Y. Suen, “Sequential combination methods for data clustering”, Journal of
computer science,2002.
[54] A. Raftery and N. Dean, “variable selection for model based clustering”, journal of American
statistical association, 101(473):168-178,2006.
[55] Pankaj K. Agarwal and Nabil H. Mustafa, “K-Means Projective clustering”, PODS Paris France,
ACM 2004.
[56] A. L. Fred and A. K. Jain., “Combining multiple clusterings using evidence accumulation”, IEEE
Transactions on Pattern Analysis and Machine Intelligence, 27(6):835–850, 2005.
[57] Ivan O.Kyrgyzov, Henri Maˆıtre and Marine Campedel, “A Method of Clustering Combination
Applied to Satellite Image Analysis”.
[58] A. P. Topchy, A. K. Jain, and W. F. Punch., “A mixture model for clustering ensembles”, In SDM,
2004.
[59] T. Lange and J.M. Buhmann, “Combining partitions by probabilistic label aggregation”, In KDD ’05:
Proceeding of theeleventh ACM SIGKDD international conference on Knowledge discovery in data
mining, pages 147–156, New York,NY, USA, 2005. ACM Press.
[60] Christian Hartvedt, “Using Context to Understand User Intentions in Image Retrieval” 2010.
[61] ZhiYong Zeng, ZhiQiang Yao, ShiGang Liu, “An Efficient and effective Image Representation for
Region-Based Image Retrieval” ICIS 2009, November.
[62] Qifeng Qiao, Peter A. Beling, “Localized Content Based Image Retrieval with Self-taught Multiple
Instance Learning” 2009 IEEE.
[63] Nidhi Singhai, Prof. Shishir K. Shandilya, “A Survey On: Content Based Image Retrieval Systems”,
International Journal of Computer Applications (0975 – 887)Volume 4 – No.2, July 2010.
[64] Steven M. Holand , “Cluster Analysis”, Department of Geology, University of Georgia, Athens, GA
30602-2501
[65] S. Guha, R. Rastogi, and K. Shim, “CURE: An efficient clustering algorithm for large databases,”
inProc. ACM SIGMOD Int. Conf. Management of Data, 1998, pp. 73–84.
[66] “ROCK: A robust clustering algorithm for categorical attributes,” Inf. Syst., vol. 25, no. 5, pp. 345–
366, 2000.
[67] G. Karypis, E. Han, and V. Kumar, “Chameleon: Hierarchical clustering using dynamic modeling,”
IEEE Computer, vol. 32, no. 8, pp. 68–75,Aug. 1999.
[68] T. Zhang, R. Ramakrishnan, and M. Livny, “BIRCH: An efficient data clustering method for very
large databases,” in Proc. ACM SIGMOD Conf. Management of Data, 1996, pp. 103–114.
[69] Ball, G., Hall, D., 1965. ISODATA, a novel method of data anlysis and pattern classification.
Technical report NTIS AD 699616. Stanford Research Institute, Stanford, CA.
[70] E. Dahlhaus, “Parallel algorithms for hierarchical clustering and applications to split decomposition
and parity graph recognition,” J. Algorithms, vol. 36, no. 2, pp. 205–240, 2000.

38
[71] C. Olson, “Parallel algorithms for hierarchical clustering,” Parallel Comput., vol. 21, pp. 1313–1325,
1995.
[72] K. Stoffel and A. Belkoniene, “Parallel K-means clustering for large data sets,” in Proc. EuroPar’99
Parallel Processing, 1999, pp.1451–1454.
[73] F. Harary, Graph Theory. Reading, MA: Addison-Wesley, 1969
[74] C.T Zahn(1971) : Graph-theoretical methods for detecting and describing clusters , IEEE trans on
Computers c(20):68-86
[75] G. Karypis; E.Han(1999) : “A hierarchical clustering Algorithm using dynamic modeling, IEEE Trans
on Computers:”, Special issue on Data analysis and Mining, 32(8):68-75
[78] J.Gramm; J.Guo(2003) : Graph modeled data clustering: fixed parameter algorithms for clique
generation, In Lecture Noted on Computer Science(LNCS),pages 109-118,Springer
[79] R. R.Shamir; D.Tsur(2002): Cluster graph modification problems. In Lecturer notes in computer
science(LNCSI),pp379-390,springer
[80] E. Hartuy ; R. Sharmir. A clustering algorithm based on graph connectivity. Information
processing,pp175-181
[81] Xiaochun Wang; D. Mitchell Wilkes (2009) : “A Divide-and-Conquer Approach for Minimum
Spanning tree-Based Clustering”, Member,IEEE Transactions on knowledge and data engineering,
Vol 21 No7
[82] Oleksandr; Grygorash(2006); Yan Zhou: “Minimum Spanning tree Based Clustering”, IEEE Tools
with artificial intelligence,pp 3-81
[83]. A. Guttman, R-tree: “A dynamic index structure for spatial searching”, in Proc. ACM SIGMOD,
1984.
[84]. T. Sellis, N. Roussopoulos, and C. Faloutsos, “The RC-tree: A dynamic index for multi-dimensional
objects”, in Proc. 12th VLDB, 1987.
[85]. D. Greene, “An implementation and performance analysis of spatial data access”, in Proc. ACM
SIGMOD,1989.
[86]. N. Beckmann, H.-P. Kriegel, R. Schneider, and B. Seeger, The R¤-tree: “An efficient and robust
access method for points and rectangles”, in Proc. ACM SIGMOD, 1990.
[87]. Y. Rui, K. Chakrabarti, S. Mehrotra, Y. Zhao, and T. S. Huang, “Dynamic clustering for optimal
retrieval in high dimensional multimedia databases”, in TR-MARS-10-97, 1997.
[88] R. O. Duda and P. E. Hart, “Pattern Classification and Scene Analysis”, Chap. 6, Wiley, New York,
1973.
[89]. H. J. Zhang and D. Zhong, “A scheme for visual feature based image retrieval”, in Proc. SPIE Storage
and Retrieval for Image and Video Databases, 1995.
[90] 3D Medical Informatics, Information Science in Multiple Dimensions chapter12 pp 333-347, Terry S.
Yoo
[91] Jayant Mishra, Anubhav Sharma and Kapil Chaturvedi, “An Unsupervised Cluster-based Image
Retrieval Algorithm using Relevance Feedback”, International Journal of Managing Information
Technology (IJMIT) Vol.3, No.2, May 2011
[92] Agrawal, Rakesh, Gehrke, Johannes, Gunopulos, Dimitrios, Raghavan, Prabhakar, 1998. “Automatic
subspace clustering of high dimensional data for data mining applications.” In: Proc. ACM SIGMOD,
pp. 94–105.
[93] Guojun Gan., “Subspace clustering for high dimensional categorical data”, May 2003. Talk Given at
SOSGSSD.
[94] Harsha S. Nagesh., “High performance subspace clustering for massive data sets.” Master's thesis,
North-western Univeristy, 2145 Sheridan Road, Evanston IL 60208, June 1999.
[95] K. G. Woo and J. H. Lee. FINDIT: “a Fast and Intelligent Subspace Clustering Algorithm using
Dimension Voting”. PhD thesis, Korea Advanced Institute of Science and Technology, Taejon,
Korea, 2002.
[96] Chun-Hung Cheng, Ada Waichee Fu, and Yi Zhang., “Entropy-based subspace clustering for mining
numerical data”, In Proceedings of the fifth ACM SIGKDD international conference on Knowledge
discovery and data mining, pages 84{93. ACM Press, 1999.

39
[97] Charles Elkan., “Using the triangle inequality to accelerate k-means”, In Tom Fawcett and Nina
Mishra, editors, ICML, pages 147–153. AAAI Press, 2003.
[98] Greg Hamerly, “Making k-means even faster”, SIAM.
[99] Lloyd, S., 1982., “Least squares quantization in PCM. IEEE Trans. Inform. Theory 28, 129–137.
Originally as an unpublished Bell laboratories Technical Note (1957).
[100] MacQueen, J., 1967, “Some methods for classification and analysis of multivariate observations”, In:
Fifth Berkeley Symposium on Mathematics. Statistics and Probability. University of California
Press, pp. 281–297.
[101] Steinhaus, H., 1956. Sur la division des corp materiels en parties. Bull. Acad. Polon.
Sci. IV (C1.III), 801–804.
[102] Zhang, Tian, Ramakrishnan, Raghu, Livny, Miron. 1996. “BIRCH: An efficient data clustering
method for very large databases.”, In: Proc.1996 ACM SIGMOD Internat. Conf. on Management of
data, vol. 25, pp. 103–114.
[103] Steinbach, Michael, Tan, Pang-Ning, Kumar, Vipin, Klooster, Steve, Potter,Christopher, 2003.,
“Discovery of climate indices using clustering”, In: Proc Ninth ACM SIGKDD Internat. Conf. on
Knowledge Discovery and Data Mining.
[104] Bradley, P.S., Fayyad, U., Reina, C., 1998., “Scaling clustering algorithms to large databases”, In:
Proc. 4th KDD.
[105] Har-peled, Sariel, Mazumdar, Soham, 2004., “Coresets for k-means and k-median clustering and
their applications”, In: Proc. 36th Annu. ACM Sympos. Theory Comput., pp. 291–300.
[106] Karypis, George, Kumar, Vipin, 1995. , “A fast and high quality multilevel scheme for partitioning
irregular graphs.”, In: Proc. Internat. Conf. on Parallel Processing, pp.113–122.
[107] Kaufman, Leonard, Rousseeuw, Peter J., 2005., Finding groups in data: An introduction to cluster
analysis. Wiley series in Probability and Statistics.
[108] Pelleg, Dan, Moore, Andrew, 2000., “X-means: Extending k-means with efficient estimation of the
number of clusters.”, In: Proc. Seventeenth Internat. Conf. on Machine Learning. pp. 727–734.
[109] Scholkopf, Bernhard, Smola, Alexander, Muller, Klaus-Robert, 1998.,“Nonlinear component
analysis as a kernel eigenvalue problem”, Neural Computer. 10 (5),1299–1319.
[110] Choras, R.S. (2007), “Image feature extraction techniques and their applications for CBIR
andbiometrics systems”, International Journal of Biology and Biomedical Engineering, Vol.1,Issue
1, Pp. 6-16.
[111] Xiaoling, W. (2009), “A Novel Circular Ring Histogram for Content-Based Image Retrieval”, First
International Workshop on Education Technology and Computer Science, Vol. 2,Pp.785-788.
[112] Wang, S. and Qin, H. (2009), “A Study of Order-Based Block Color Feature Image Retrieval
Compared with Cumulative Color Histogram Method”, Sixth Internationa lConference on Fuzzy
Systems and Knowledge Discovery, FSKD '09, IEEE Xplore, Vol. 1,Pp. 81 – 84.

A Survey On: Content Based Image Retrieval Systems Using Clustering Techniques For Large Data sets

More Related Content

What's hot (16)

Similar to A Survey On: Content Based Image Retrieval Systems Using Clustering Techniques For Large Data sets (20)

More from IJMIT JOURNAL (20)

Recently uploaded (20)

A Survey On: Content Based Image Retrieval Systems Using Clustering Techniques For Large Data sets