International Journal of Engineering Trends and Technology (IJETT) – Volume 15 Number 3 – Sep 2014
Region Based Classification Algorithm for
Medical Image Databases
1
R.Krishnam Raju, 2B.Jyothi, 3Y.Madhavee Latha
1
2
Department of ECE, Mallareddy college of Engineering& Technology, Hyderabad, India
Associate Professor Department of ECE, Mallareddy college of Engineering& Technology, Hyderabad, India
3
Professor Department of ECE Malla Reddy Engineering College for Women, Hyderabad, India
Abstract
In this paper presents the region based
classification algorithm for medical image
database. It provides better results and efficient
retrieval of medical images. The proposed method
follows three steps. Firstly, segment the medical in
to regions based on edge following technique. This
technique is more accurate and efficiency
compare to existing contour methods. Secondly,
calculate the feature vectors of each region using
signature. Thirdly, calculate the distance between
query feature and database image features. The
efficiency and performance of the presented
method has been evaluated using a dataset of
about 500 simulated. The experimental results
show that the proposed method has more
efficiently and effectively for medical applications
in CBIR systems.
Index Term: - Content-based image retrieval,
medical image databases, region matching
I.INTRODUCTION
The explosive growth of the internet and the
wide use of digital content necessitate the
development of effective ways of managing the
visual information by its content and have increased
the need for efficient image retrieval procedure [1].
Due to the rapid development of computing
hardware, digital acquisition of information has
become one popular method in recent years. Every
day, G-bytes of images are generated by both military
and civilian equipment. Large set of medical images,
architectural and engineering designs, journalism and
advertising, are worth mentioning. Consequently,
how to make use of this huge amount of images
effectively becomes a highly challenging problem
[2]. As a result, studies on Content Based Image
Retrieval (CBIR) have emerged and have been an
active research from the past decade.
In most of the image retrieval systems, a query is
specified by an image to be matched. We refer to this
as an overall search, as similarity is based on the
ISSN: 2231-5381
overall properties of images. By contrast, there are
also partial search querying systems that retrieve
based on a particular region in an image. A contentbased image retrieval method named CBDIR,
segmenting the teeth of dental study models (plaster
casts of the dentition), exhibits varieties of
malocclusions [3]. Medical and general purpose
image retrieval (MGIR) method is used for retrieving
medical and general purpose images from databases,
robust to scaling and translation of objects within an
image [4]. The Colorimetric-Based Retardation
Measurement Method (CBRM) is a method in which
each image is first decomposed into regions [5]. A
measure for the overall similarity between images is
developed using a region-matching scheme that
integrates the properties of all the regions in the
images [6].
The CBMIR system is interactive and the
user is allowed to correct the results of the segmented
images. The user can identify and extract interesting
images or regions from all segmented images. The
user can even specify the class to which an image
belongs. Based on the properties of individual
regions and spatial relationships between such
regions, the CBMIR system takes the responsibility
of efficient storage, representation, and retrieval of
images.
In this paper is organized as follows.
Proposed method in section II. The simulation results
are presented in Section III. Concluding remarks are
made in Section IV.
II. PROPOSED METHOD
The proposed method has better retrieval
accuracy and efficient method. In this work, retrieval
the medical image based on region based
segmentation. To segment the image into regions or
dominant objects based edge following technique.
After that apply the label of each region. The block
diagram of proposed method as shown in figure.1.
http://www.ijettjournal.org
Page 141
International Journal of Engineering Trends and Technology (IJETT) – Volume 15 Number 3 – Sep 2014
DB
image
s
Edge
following
Where
Feature
selection
K=max(sqrt(Mx+My))
Similarity
measure
Each component is the convolution between the
image and the corresponding difference mask, i.e.
Mx=-Gy*f
Query
image
Edge
following
Feature
selection
My=Gx+f
Relevance
image
Fig.1. block diagram of proposed method
The algorithm of proposed method is given by
for k=1: DBIMAGES
read image k;
SEG=segment the image using edge following
contour
[B,pk]=Bwlabel(seg);
for i=1: Pk %pk is number of components in kth
image
kk=B(i) %read the ith component of kth image
find Si; */ Si is the signature of component i
calculate Fi1, Fi2… Fi8
save k, i, Fi1, Fi2 …Fi8
*/ save ten values for component i, in database Dv
end;
end
query matching
input Q; */Q is the query image, SOM segmented
find Fq;*/ calculate the feature vector of query image
Q
for J=1: N */ N is the number of images in database
Dd
compare Q; */ search for the most similar images in
database Dy
*/ (compare number of components and feature
vectors)
report similarities; */ report partially or completely
similar cases.
end;
where Gx and Gy are the difference masks of the
Gaussian weighted image moment vector operator in
the x and y directions.
At the pixel position (i, j) of an image, the successive
positions of the edges are then calculated by a 3 ×
3matrix
L(r,c) = αM(r,c) + βD(r,c) +€E(r,c)
Where α, β, and ε are the weight parameters that
control the edge to flow around an object.
The algorithm of edge following technique as shown
below
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
A. Edge following technique
14.
It is Law’s texture and Canny edge
detection. Edge following technique can be divided
into the two steps. They are edge map and average
vector field. In law’s texture, calculate a convolving
an input image with each of the masks. In average
vector field, the edge vector field is calculated
according to the following equations:
15.
Read the image
Generate the Gaussian mask
Convolution between image and Gaussian
mask
Apply gradient %initial gradient
[u n]=Initializing the contour position
For 1: no_of contours
[nx,ny]=gradient(u);
%%%%%%%%%%edge
vector
field
convolution%%%%%%%%%%%%%%%
%
Mx=sqrt(Nx2+Ny2);
Ang=angle(u);
[N]=gradiend(nx,ny);
%%%%%%%law
texture .here we donot calculate convolution
between image and mask. Because
convolution applied in step 3
P=gradient(nx,ny);%%%%%%%%%%%%
edge detecton
E=N+p;%%%%%%%%edge
map=texture+ edge detecton
%%%%%%%%%%%%%end
of
edge
map%%%%%%%%%%%%%%%%%%%
%%%
u=u+(Ω*Mx+ α*E+β*ang); %%%%%edge
following technique or contour update
End of contour
E=1/k (Mx+My)
ISSN: 2231-5381
http://www.ijettjournal.org
Page 142
International Journal of Engineering Trends and Technology (IJETT) – Volume 15 Number 3 – Sep 2014
B. Feature selection
1
In this work, we take the signature as a
feature vector. Features are extracting from database
images and query image. In signature feature
calculated based on three ratios. They are 1)
Signature Height Width Ratio 2) Signature
Occupancy Ratio.
The signature height width is ratio obtained
by dividing signature height to signature width. The
height is the maximum length of the object obtained
from the segmented region from medical image.
Similarly the width is also calculated considering the
row of maximum length.
The signature occupancy ratio is the ratio of
number of pixels which belong to the signature to the
total pixels in the signature image and is given by
7
Fig.2. query image number 1
1
6
5
9
11
4
7
12
10
8
2
3
13
23
24
14
22
17
18
15
31
16
21
41
27
20
30
19
28
C. Distance calculation
Euclidean distance is the distance between
energy of images from database and energy of query
image. The formula of Euclidean distance is given by
dis tan ce  query 2  dbimages 2
29
Fig.3. Retrive images for a sample query of 1 in the
database using proposed method
Here
Query is the energy of query image and db images is
the energy of database images.
401
III. EXPERIMENTAL RESULTS
A dataset of about 500 simulated, but
realistic computed tomography and magnetic
resonance images (MRI) is used. These work
simulated using MATLAB TOOL.
418
Fig.4. query image number 401
IV. CONCLUSION
In this paper employs novel technique in which each
image is first decomposed into regions and then
calculate theg the signatures and computing feature
vectors. It is more accuracy and retrieval effiency.
The results show the retrieval images more relevant
to query image.
ISSN: 2231-5381
http://www.ijettjournal.org
Page 143
International Journal of Engineering Trends and Technology (IJETT) – Volume 15 Number 3 – Sep 2014
401
416
418
413
424
429
425
430
449
407
444
426
415
446
422
423
440
439
414
443
441
404
406
419
442
438
448
437
REFERENCES
[1] S. Liapis, G. Tziritas, “Color and texture image retrieval using
chromaticity histograms and wavelet frames,” IEEE Transactions
on Multimedia, Vol. 6, Oct. 2004.
[2] A. Mumtaz, S. A. M. Gilani, T. Jameel, “A Novel Texture
Image Retrieval System based on Dual Tree Complex Wavelet
Transform and Support Vector Machines,” IEEE 2nd International
Conference on Emerging Technologies, 2006.
[3]. E. G. Petrakis, and C. Faloutsos. Similarity searching in
medical image databases. IEEE Trans Knowl Data Eng 9, pp. 43547, 1997.
[4]. A. H. Pilevar, M. Sukumar, A. R. Gowda, and E. T. Roy.
CBDIR: Content-based dental image retrieval. J Indian Orthod Soc
6, pp. 38-52, 2005.
412
445
[5]. A. H. Pilevar, and M. T. Pilevar. MGIR: An image retrieval
method for medical and general image databases. 11th Joint
Conference on Information Science (JCIS 2008), Dec. 15-20,
2008.
Fig.5. Retrive images for a sample query of 401 in the
database using proposed method
[6]. A. H. Pilevar, and M. T. Pilevar. CBRM: A new content-based
regional-matching image retrieval method. Proceeding of the
eighth IASTED international conference, visualization, imaging
and image processing, Sept.1-3, 2008.
251
289
Fig.6. query image number 251
251
299
298
297
296
286
289
291
300
288
290
285
294
282
287
295
281
292
293
284
280
283
258
259
257
279
276
252
267
278
Fig.7. Retrive images for a sample query of 401 in the
database using proposed method
ISSN: 2231-5381
http://www.ijettjournal.org
Page 144