Mri brain image retrieval using multi support vector machine classifier

International Journal of Advanced Information Science and Technology (IJAIST) ISSN: 2319:2682
Vol.10, No.10, February 2013
29
Abstract-Content of image retrieval is the process of finding
relevant image from large collection of image database using
visual queries. Medical images have led to growth in large image
collection. To enhance the medical image retrieval for
diagnostics, research and teaching purposes is done by CBIR.
The content of medical images is difficult to describe in words or
textual form. The proposed system uses multiple image queries
for finding desired images from the database. The system
performance is improved by the multiple image queries instead
of single image query. Pre-processing of the query image is done
by median filter to remove the noise. Then the filtered image is
given as input to the feature extraction technique which is a
transformation of input image into set of features such as texture
and shape. Feature extraction is done by the Gray level co-
occurrence matrix algorithm that contains information about the
position of pixels having similar gray level values. The feature
optimization is done on the extracted features to select best
features out of it to train the classifier. SVM (Support Vector
machine) classifier is to group items that have similar feature
values into three categories such as normal, benign and
malignant. Then SVM classifier is followed by KNN (K-nearest
neighbor) which search the corresponding database index will be
computed by similarity feature matching. The query image is
classified by the classifier to a particular class and the relevant
images are retrieved from the database.
Index Terms- GLCM, Image retrieval, Feature Extraction,
MRI, SVM classifier.
I. INTRODUCTION
Among the applications of computer science to the
field of medicine, the processing of medical image data is
playing an increasingly important role. With medical imaging
techniques such as X-Ray, computer tomography, magnetic
resonance imaging, and ultrasound, the amount of digital
images that are produced in hospitals is increasing incredibly
fast. Thus, the tasks of efficiently storing, processing and
retrieving medical image data have become important
research topics. Many hospitals use picture archiving and
communication systems (PACS), which are basically
computer networks that are used for storage, retrieval, and
distribution of medical image data. In such a network, a
central server provides access to an image database from
which clients such as medical staff can retrieve images by
using metadata like the name of the patient, the date, the
imaging method, the body part, etc.
Metadata-based retrieval is done via standard
database tasks that are relatively easy to implement. If the
metadata for an image is not sufficient to formulate a precise
enough query, a textual query can be given. This query can
for example be a set of keywords or a full textual description
of the desired images. Then, the PACS searches for database
images with similar descriptions. This retrieval task is more
complicated and involves techniques from the field of text
retrieval. If for example a doctor wants to compare X-ray
images of his current patient with images from similar cases,
he could also use these images as queries and let the PACS
find the most similar entries in the database. This kind of
searching for images is called content-based image retrieval
(CBIR) and is currently part of the research of many computer
science groups, who are trying to find models for the
similarity of digital images. Several content based image
retrieval systems are currently being developed.
The applications of computer vision techniques
present an image retrieval problem which is explicated as the
problem of searching for digital images in large databases. An
image retrieval system is a computer system for searching and
retrieving images from a large database of digital images.
Color, Shape and texture are important cue in extracting
information from images; these histograms are widely used in
content based image retrieval (Serge Belongie et al, 1998).
Color and texture contain important information for two
images with similar color histograms that can represent very
different things. Therefore the use of shape-describing
features is essential in an efficient content-based image
retrieval system. Although shape description has been
intensively researched, there exists no direct answer as to
what kind of shape features is incorporated into such a system
(Jorma Laaksonen et al, 2000).
Most traditional and common methods of image
retrieval utilize some technique of adding metadata to the
images so that retrieval can be performed over the annotation
words. The effectiveness of content-based image retrieval
systems can be improved by combining image features or by
considering image similarities, as computed from multiple
feature vectors (Ricardo da S. Torresa et al, 2009). A retrieval
scheme is proposed making use of local color invariant
information in order to produce semi global shape invariants,
to obtain a viewpoint invariant, to have a high- dimensional
object descriptor and to be used as an index for discriminatory
MRI Brain Image Retrieval Using Multi Support
Vector Machine Classifier
R. Guruvasuki
PG Scholar in
M.E Computer & Communication Engineering,
M.A.M College of Engineering, Trichy, India.
guruvasuki@gmail.com
A. Josephine Pushpa Arasi M.E.,
Associate Professor/Department of IT
M.A.M College of Engineering, Trichy,
India.
josephineit@mamce.org

30
image retrieval (Theo Gevers and Arnold W. M.
Smeulders,2000).It is aimed at efficient retrieval of relevant
images from large image database based on imagery features;
these are typically extracted from visual properties of query
image in the database. The relevance between a query image
and database image is ranked based on similarity measure
computed from the features (Yixin chen and James Z. Wang,
2002).
The input images are analyzed to extract the features
and these features are stored in the image database, along with
the original images. These features could be shape features,
texture features or color features. Whenever an image is
submitted for search, it is analyzed and its features are
extracted (Babu M. Mehtre et al, 1997). The easiest way to
use regional features is to use blocks of fixed size and
location, which is called as partitioning of the image for local
feature extraction. These blocks do not take into account any
semantics of the image itself (Henning Muller et al, 2003).
Region based systems typically use moment
descriptors that include Geometrical moments, Zernike
moments and Legendre moments. The Zernike basis functions
satisfy the orthogonal property, implying that the contribution
of each moment coefficient to the underlying image is unique
and independent. Here no redundant information is
overlapped between the moments (Chia-Hung Wei et al,
2009). The Zernike moments cannot capture spectral feature
in radial directions. Its moments cannot examine shape details
in it (Dengsheng Zhang and Guojun Lu , 2002). Boundary-
based systems use the contour of the objects and usually give
better results for images that are distinguishable by their
contours (B.Sathyabama, 2011). It allows nonlinear
combination of image similarities and is validated through
several experiments, where the images are retrieved based on
the shape of their objects.
Figure 1. General flowchart for the CBIR system for a given query image or
images.
In CBIR systems, images are typically represented
with feature vectors extracted using low-level image
processing techniques. However, similarities in feature vector
level do not always guarantee the semantic similarity (i.e.,
interpretations of images according to their predefined
categories) between query image and retrieved images. This is
known as the semantic gap problem.
In this paper, we will explore the effect of slide-level
retrieval system with multiple query images in order to
increase the semantic relevance of query image set and
retrieved images. A general flowchart of the proposed CBIR
system is illustrated in Figure 1. It shows the main steps of the
CBIR algorithm, e.g., feature extraction, major disease-type
classification (first tier), and image retrieval according to the
subtypes of the diseases (second tier).
II. Background and Related Work
Most of the commercial search engines (e.g., Google,
Yahoo!, Bing Image Search) are built around a semantic
search, i.e., the user needs to type in a series of keywords and
the images in those databases are also annotated using
keywords; the match is accomplished primarily through these
keywords. CBIR systems have been developed in the recent
years to organize and utilize the valuable image sources
effectively and efficiently for diverse collections of images.
Most of the recent CBIR systems in biomedicine are designed
to classify and retrieve images according to the anatomical
categories of their content, i.e., head or chest X-ray images or
abdominal CT images.
Azhar Quddus and Otman Basir (2012) propose a
novel technique for subject identification and semantic
classification of brain images. They proposed a technique that
associates the query slices with a specific area of the brain. It
identifies the 3-D volume of the patient based on the query
slice. The identification and classification are achieved using
features in the multi-scale wavelet domain. The limitation of
the proposed retrieval technique is that slice images should
have the same resolution and dimensions. The retrieval in
sagittal view is not possible within the semantic regions.
Shen-Tat Goh and Kian-Lee Tan (2000) present an
image retrieval system that is based on a set of clusters from
the image. The multiple features are the color, the size and the
spatial location of the cluster. They have also proposed an
index structure for the speedy retrieval of images. Without
texture and shape features it is insufficient to fully represent
the content of an image.
Heng Qi, KeqiuLi, YanmingShen and WenyuQu
(2010) propose a feature matching strategy to compute the
dissimilarity value among the feature vectors extracted from
images. Finally, they have combined the shape description
method and the feature matching strategy. They have also
conducted also experiments on a standard image set to
evaluate their solution.
Sami Brandt, Jorma Laaksonen and Erkki Oja (2000)
studied content-based image retrieval with shape describing
features. They used edges of non-segmented images for
feature vectors. The results were obtained from decimated
magnitude spectrum of the image using features including the
edge-histogram and the co-occurrence matrix of the edge

31
directions. Performance evolution of the retrieval process is
not discussed.
Pourghassem and Ghassemian (2008) proposed a
two-level hierarchical medical image classification method.
The first level was used to classify the images into the merged
and non-merged classes. They tested their algorithm on
medical X-ray images of 40 classes. Although this is a two-
level hierarchical classification, it is different from our
approach because only the merged classes were evaluated in
the second level to be classified with multilayer perceptron
(MLP) classifiers into 1 of 40 classes.
Hatice Akakin and Metin N. Gurcan (2012) have
proposed CBIR system which uses a multi-tiered approach for
retrieving images. This system supports multi-image query
and slide-level image retrieval. New weighting terms are
defined for multiple-image query and retrieval. Here they
have not discussed the performance evolution of the retrieval
process.
Dimitris K. Iakovidis et al (2009) have proposed a
scheme which involves block-based feature extraction from
images following the clustering of the feature space to form
higher level, there by having semantically meaningful
patterns. The expectation maximization algorithm uses an
iterative approach to automatically determine the number of
clusters. Then, the similarity between two clusters is
estimated. Experiments were conducted on a large set of
images.
Weidong Cai, (David) Dagan Feng, and Roger
Fulton (2000) presented a positron emission tomography and
content based image retrieval system. This system enables to
reduce storage requirements. This method is enabled to handle
large number of patient data. It also offers advantages in
medical image data management.
Jun Yue, Zhenbo Li, Lu Liu and Zetian Fu (2011)
have presented a method to extract color and texture features
using content-based image retrieval. Color and texture
features based on a co-occurrence matrix are extracted to form
feature vectors. Then the characteristics of the global, local
color histograms and texture features are compared and
analyzed. CBIR system is designed using color and texture
combined features by constructing weights of feature vectors.
In addition they have discussed the performance measure of
the system.
III. OVERVIEW OF PROPOSED SYSTEM
ARCHITECTURE
We propose a new methodology for content based
medical image retrieval based on multi image query system.
The given user query image is projected onto the feature
space by extracting the texture features from co-occurrence
matrix constructed from the original image matrix.
Categorization of query image is done by means of SVM
classifier. Meanwhile the database images are also classified
in the same way for refining the searching. KNN search
algorithm is utilized for searching the relevant images for the
given query image. This system is suitable for retrieving the
similar images of multi-image queries also. Our methods
supports
 Yielding the good performance by considering single
modality(MRI)
 Enables multi image query instead of one single
image.
 Easy to extend the framework for various diseases.
The feature extraction is a technique for capturing
visual content of images for indexing and retrieval. Feature
extraction is defined as tracing those pixels in an image that
have some precise characteristics. They ought to be easy to
compute for the approach to be possible for a large image
collection and rapid retrieval. Since the users will finally
decide the aptness of the retrieved images they should relate
well with the human perceptual personality. This process is
otherwise known as Content Based Image Indexing (CBII).
A. System Architecture:
Figure 2. Proposed CBIR system architecture.
B. Preprocessing
We have seen that smoothing (low pass) filters
reduce noise. However, the underlying assumption is that the
neighboring pixels represent additional samples of the same
value as the reference pixel, i.e. they represent the same
feature. At edges, this is clearly not true, and blurring of
features results. You have used convolution techniques to
implement weighting kernels as a neighborhood function,
which represented a linear process. There are also nonlinear
neighborhood operations that can be performed for the
purpose of noise reduction that can do a better job of
preserving edges than simple smoothing filters. In the median
filtering operation, the pixel values in the neighborhood
window are ranked according to intensity, and the middle
value (the median) becomes the output value for the pixel
Preprocessing
Database
Feature
extraction
Multi-SVM
classifier
Classified
feature matrix
KNN search
Retrieval
result
Query image
Preprocessing
Feature
extraction
Feature vector
Multi-SVM
classifier
Feature
vector

32
under evaluation. Median filtering does not shift boundaries,
as can happen with conventional smoothing filters. Since the
median is less sensitive than the mean to extreme values
(outliers), those extreme values are more effectively removed.
Median filtering preserves the edges.
 No reduction in contrast across steps, since output
values available consist only of those present in the
neighborhood (no averages).
 Median filtering does not shift boundaries, as can
happen with conventional smoothing filters (a
contrast dependent problem).
 Since the median is less sensitive than the mean to
extreme values (outliers), those extreme values are
more effectively removed.
Figure 3. Median Filtering value
The median is, in a sense, a more robust “average”
than the mean, as it is not affected by outliers (extreme
values). The output pixel value is one of the neighboring
values; new “unrealistic” values are not created near edges.
The edges are minimally degraded; median filters can be
applied repeatedly, if necessary
C. Texture feature extraction
When the input data to an algorithm is too large to be
processed and it is suspected to be notoriously redundant
(much data, but not much information) then the input data
will be transformed into a reduced representation set of
features (also named features vector). Transforming the input
data into the set of features is called feature extraction. The
features provide the characteristics of the input type to the
classifier by considering the description of the relevant
properties of the image into a feature space. If the features
extracted are carefully chosen, it is expected that they will
extract the relevant information from the input data in order to
perform the desired task using this reduced representation
instead of the full size input. Feature extraction involves
simplifying the amount of resources required to describe a
large set of data accurately.
When performing analysis of complex data one of
the major problems stems from the number of variables
involved. Analysis with a large number of variables generally
requires a large amount of memory and computation power or
a classification algorithm which over fits the training sample
and generalizes poorly to new samples. Feature extraction can
be used in the area of image processing which involves using
algorithms to detect and isolate various desired portions or
shapes (features) of a digitized image or video stream.
Another important feature processing stage is feature
selection. This section has given an overview of many
possible features that can be used, and we would like to use a
feature set that encodes a diverse variety of types of
information in order to enhance discrimination between
normal and abnormal areas.
However, when large and complicated feature sets
are used to train on smaller training sets, classifiers can
‘overfit’ the learned model, since it is likely that spurious
patterns can be found that can accurately classify the training
data, but are not relevant to unseen test data. Feature selection
is partially up to the designer to select an appropriate feature
set, but automatic methods can also be used. In selecting
features, it is important to consider whether features will help
in discriminating unseen data, and how complicated the
interactions between the features are likely to be in order for
them to be used in discrimination.
D. Gray Level Co-occurrence Matrix
 A gray level co-occurrence matrix (GLCM) contains
information about the positions of pixels having
similar gray level values.
 A co-occurrence matrix is a two-dimensional array,
P in which both the rows and the columns represent
a set of possible image values.
 A GLCM Pd[i,j] is defined by first specifying a
displacement vector d=(dx,dy) and counting all pairs
of pixels separated by d having gray levels i and j.
 The GLCM is defined by
Where nij is the number of occurrences of the pixel
values (i, j) lying at distance d in the image. The co-
occurrence matrix Pd has dimension n × n, where n is the
number of gray levels in the image. For example, if d=
(1,1)
Figure 4. Extraction by GLCM
There are 16 pairs of pixels in the image which
satisfy this spatial separation. Since there are only three gray
levels, P[i,j] is a 3×3 matrix.
Algorithm for GLCM

33
 Count all pairs of pixels in which the first pixel has a
value i, and its matching pair displaced from the first
pixel by d has a value of j.
 This count is entered in the ith row and jth column of
the matrix Pd[i,j]
 Note that Pd[i,j] is not symmetric, since the number
of pairs of pixels having gray levels [i,j] does not
necessarily equal the number of pixel pairs having
gray levels [j,i].
E. Support Vector Machine classifier
Aim of SVM classifier is to group items that have
similar feature values into groups. Classifier achieves this by
making a classification decision based on the value of the
linear combination of the features.
 data setup: our dataset contains three classes as
normal, benign (non-cancerous), malignant
(cancerous) each N samples. The data is 2D plot
original data for visual inspection.
 SVM with linear kernel (-t 0). We want to find the
best parameter value C using 2-fold cross validation
(meaning use 1/2 data to train, the other 1/2 to test).
 After finding the best parameter value for C, we train
the entire data again using this parameter value.
 plot support vectors.
 plot decision area.
Figure 5. Support Vector Machine classifier
Expression for hyper plane: w.x+b = 0
Margin is d1+d2.
where x – Set of training vectors
w–vectors perpendicular to the separating hyper plane
b–offset parameter which allows the increase of the margin
A grouping of all the classes in two disjoints groups
of classes. This grouping is then used to train a SVM
classifier in the root node of the decision tree, using the
samples of the first group as positive examples and the
samples of the second group as negative examples. The
classes from the first clustering group are being assigned to
the first (left) subtree, while the classes of the second
clustering group are being assigned to the (right) second
subtree. The process continues recursively until there is only
one class per group which defines a leaf in the decision tree.
Figure 6. n-class SVM
F. Relevant Image search by KNN within a category
 Place items in class to which they are “closest”.
 Must determine distance between an item and a
class.
 Sorting the distances of most number of close images
within a particular class and the nearest neighbor
indexes are computed.
G. Similarity Measurements and retrieval
After getting the relevant image ids from KNN
search the corresponding database index will be computed by
similarity feature matching. With the help of that database
index values the relevant images are retrieved and displayed.
The distance values are displayed and plotted as a bar graph.
Figure 7. Graph for distance values.
IV. PERFORMANCE EVALUATION
A. Precision & Recall Analysis
class A-Normal.
class B-Benign
class C-Malignant

34
CBIR performance is analyzed by computing the
values of precision and recall.
Precision = Number of relevant images retrieved / Total
number of images retrieved
Figure 8. Precision Graph
Recall = Number of relevant images retrieved / Total number
of relevant images in the database.
Figure 9.Recall Graph
Retrieval Time is analyzed by using number of query images.
Figure 10. Retrieval Time Graph
B. Results & Discussion
Multi- image Query /single image Query of any
medical image (e.g. brain, liver) is preprocessed and gets the
relevant image from the database and also gets information of
the image. Information of the image contains database ID,
distance value between query and relevant image, category of
the image such as normal, benign, and malignant.
Figure 11 (a) (b) Selection of multiple query images for retrieval.
Figure 12.Retrieved Images for the given query image.

35
Figure 13. Distance measure and classification category for the given query
image with relevant images.
V CONCLUSION
We concluded that Gray Level Co-occurrence Matrix
is used for feature extraction and Multi-Support Vector
Machine (M-SVM) classifier is used for medical image
classification. Then the system performance is improved by
the multiple image queries than single image query. Then the
challenges are to reduce the search space of query images by
categorizing the images using multiclass support vector
machines (M-SVM) and to overcome the limitation on the
specification of image content of single image queries.
REFERENCES
[1]. Hatice Cinar Akakin and Metin N. Gurcan,” Content-Based
Microscopic Image Retrieval System for Multi-Image
Queries,” IEEE Transaction on Information Technology in
Biomedicine, vol. 16, No. 4, July 2012.
[2]. M. M. Rahman, P. Bhattacharya, and B. C. Desai, “A
framework for medical image retrieval using machine
learning and statistical similarity matching techniques with
relevance feedback”, IEEE Trans. Inf. Technol.Biomed.,
vol. 11, no. 1, pp. 58–69, Jan. 2007.
[3]. L. Yang, R. Jin, L. Mummert, R. Sukthankar, A. Goode, B.
Zheng, S.C. H. Hoi, and M. Satyanarayanan, “A boosting
framework for visuality preserving distance metric learning
and its application to medical image retrieval”, IEEE Trans.
Pattern Anal. Mach. Intell., vol. 32, no. 1, pp. 30–44, Jan.
2010.
[4]. C.-W.Hsu andC.-J. Lin, “A comparison of methods formulti-
class support vector machines,” IEEE Trans. Neural Netw.,
vol. 13, no. 2, pp. 415–425, Mar. 2002.
[5]. Q. Iqbal and J. K. Aggarwal, “Feature integration, multi-
image queries and relevance feedback in image retrieval,”
in Proc. 6th Int. Conf. Vis. Inf.Syst. DMS, 2003, pp. 467–
474.
[6]. Dipankar Hazra (2011),”Texture Recognition with
combined GLCM,Wavelet and Rotated Wavelet Features,”
International Journal of Computer and Electrical
Engineering, Vol.3, No.1.
[7]. El-Sayed A and El-Dahshan (2009), “ A hybrid technique
for automatic MRI brain images classification,”
International Journal of Computer Application, vol 2, no.3.
[8]. Evangelia I. Zacharaki and Sumei Wang (2009), “MRI-
Based Classification of Brain Tumor Type and Grade
using SVMRFE,” IEEE Transactions on Medical Imaging,
Vol.6 No.10.
[9]. Hatice Cinar Akakin and Metin N. Gurcan (2012),”
Content-Based Microscopic Image Retrieval System for
Multi-Image Queries,” IEEE Transaction on Information
Technology in Biomedicine, vol. 16, no. 4.
[10]. Iakovidis D., Pelekis N. , Kotsifakos E. ,Kopanakis I.,
Karanikas H. , and Theodoridis Y. (2009),”A pattern
similarity scheme for medical image retrieval,” IEEE Trans.
Inf. Technol. Biomed., vol. 13, no. 4, pp. 442–450.
[11]. Jagannathan N.R. and Sandeep Chaplot, Patnaik L.M.
(2006), “Classification of Magnetic Resonance Brain
Images using wavelets as input to Support Vector Machine
and neural network,” ELSEVIER, Biomedical Signal
Processing and Control, vol.1.
[12]. W. Hsu, L. R. Long, and S. Antani, “Spirs: A framework for
content-based image retrieval from large biomedical
databases,” Stud. Health Technol. Informat., vol. 129, no.
Pt. 1, pp. 188–192, 2007.
[13]. L. Zheng, A. Wetzel, J. Gilbertson, and M. Becich, “Design
and analysis of a content-based pathology image retrieval
system,” IEEE Trans. Inf. Technol. Biomed., vol. 7, no. 4,
pp. 249–255, Dec. 2003.
[14]. H. Pourghassem and H. Ghassemian, “Content-based
medical image classification using a new hierarchical
merging scheme,” Comput. Med. Imag. Graph., vol. 32,
no. 8, pp. 651–661, 2008. Mar. 2002.
[15]. G.-H. Liu, L. Zhang, Y.-K. Hou, Z.-Y. Li, and J.-Y. Yang,
“Image retrieval based on multi-text on histogram,” Pattern
Recognit., vol. 43, pp. 2380–2389, Jul. 2010.
[16]. D. Iakovidis, N. Pelekis, E. Kotsifakos, I. Kopanakis, H.
Karanikas, and Y. Theodoridis, “A pattern similarity cheme
for medical image retrieval,” IEEE Trans. Inf. Technol.
Biomed., vol. 13, no. 4, pp. 442–450, Jul. 2009.
[17]. E.A. El-Dahshan, T. Hosny, A. B. M.Salem “Hybrid
intelligent techniques for MRI brain images classification”,
Digital Signal Processing, vol.20(2), pp. 433 – 441, 2010.
[18]. K. Kasiri, K. Kazemi, M. J. Dehghani, M.S. Helfroush,
“Atlas - based segmentation of brain MR images using
least square support vector machines”, Image Processing
Theory, Tools and Applications, pp. 306-310, 2010.
[19]. M. M. Rahman, P. Bhattacharya, and B. C. Desai, “A
unified image retrieval framework on local visual and
semantic concept-based feature spaces,” J. Vis. Commun.
Image Represent., vol. 20, pp. 450–462, 2009.
[20]. D. Unay, A. Ekin, and R. Jasinschi, “Local structure-based
region-of interest retrieval in brain MR images,” IEEE
Trans. Inf. Technol. Biomed., vol. 14, no. 4, pp. 897–903,
Jul. 2010.
[21]. G.Quellec, M. Lamard,G. Cazuguel, B. Cochener, and C.
Roux, “Wavelet optimization for content-based image

36
retrieval in medical databases,”Med. Image Anal., vol. 14,
pp. 227–241, 2010.
R. Guruvasuki received her B.Tech
degree in Electronics and Communication
Engineering from Kalasalingam
University, Srivilliputtur, Tamilnadu,
India in 2011. She is currently pursuing
her M.E degree in Computer and
Communication Engineering from
M.A.M. College of Engineering, Trichy,
Anna University, Chennai, Tamilnadu,
India.Her research interest include
Medical Image Processing, Digital Signal
Processing and Mobile communication.
Mrs. A. Josephine Pushpa Arasi received
her B.E degree in Electronics and
Communication Engineering from
Dr.Paul’s Engineering college, vannur,
Madras University, Chennai, Tamilnadu,
India in 2004 and received her M.E
degree in Applied Electronics from
Arunai Engineering college,
Thiruvannamalai, Anna University,
Chennai, Tamilnadu, India in 2006. She
has been currently working as Associate
Professor, Department of Information
Technology at M.A.M College of
Engineering, Trichy, Tamilnadu,
India.Her research interest include Digital
Signal Processing, Digital Image
Processing and VLSI.

Mri brain image retrieval using multi support vector machine classifier

More Related Content

What's hot (19)

Similar to Mri brain image retrieval using multi support vector machine classifier (20)

More from srilaxmi524 (8)

Recently uploaded (20)

Mri brain image retrieval using multi support vector machine classifier