Sunteți pe pagina 1din 4

68 (IJCNS) International Journal of Computer and Network Security,

Vol. 2, No. 3, March 2010

DDOM: The Dynamic Data Oriented Model for


Image Using Binary Tree
A.Shahrzad Khashandarag1*, A.Mousavi2, R.Aliabadian3, D. Kheirandish4 and A.Ranjide Rezai5
1
Young Researchers Club of Tabriz, Islamic Azad University Tabriz Branch,
Tabriz-Iran
1
a.shahrzad@iaut.ac.ir
2 3 4 5
alireza.mousavi499@gmail.com ramin.ali@gmail.com kheyrandish@iaupmogan.ac.ir aranboy@gmail.com

GIF (Graphics Interchange Format) is an 8-bit-per-pixel


Abstract: This paper presents a dynamic data oriented
model (DDOM) of image. The ability of model is very fast image bitmap image format that was introduced by CompuServe in
processing against Data oriented Model that Habibizad and 1987 and has since come into widespread usage on the
cooperators proposed in [8]. In our approach, the Sobel World Wide Web due to its wide support and portability.
algorithm is used for edge detection of image and histogram
thresholding is used for clustering of image and binary tree is The format uses a palette of up to 256 distinct colors from
used for anatomy of model. Each node of this tree represents the the 24-bit RGB color space. GIF images are compressed
feature of image or sub image. By using the presented model in using the LZW lossless data compression technique to
this paper, tasks as measuring similarity ratio, segmentation, reduce the file size without degrading the visual quality
clustering … can be done with high desired precision and [10].
corresponding speed [8].
Keywords: Data oriented modeling; Image Segmentation; In computing, JPEG (pronounced JAY-peg) is a commonly
histogram thresholding; Binary Tree. used standard method of compression for photographic
images. The name JPEG stands for Joint Photographic
1. Introduction Experts Group, the name of the committee who created the
Recently many researchers have been studied on data standard. The group was organized in 1986, issuing a
structures for image processing tasks. In [1-7] has been standard in 1992, which was approved in 1994 as ISO
explained the data structures of images. The main idea of 10918-1. The compression method is usually lossy
this paper is to present an optimized and dynamic data compression, meaning that some visual quality is lost in the
oriented model of image. Dynamic Data Oriented Modeling process, although there are variations on the standard base
(DDOM) is an approach which models concepts by using line JPEG, which are lossless. There is also an interlaced
the data structure. We have introduced dynamic data "progressive" format, in which data is compressed in
oriented modeling of image for fast image processing. The multiple passes of progressively higher detail. This is ideal
Conventional methods use the color of the pixels to model for large images that will be displayed whilst downloading
the image and try to reduce the size of image (Compression) over a slow connection, allowing a reasonable preview
[8]. Previous image models are: before all the data has been retrieved [11].

• BMP DOM (Data Oriented Model) proposed by Habibizad and


• JPEG cooperators. This model is design for tasks that are very fast.
In the related works section, this model will be explained.
• GIF
• DOM
2. Related Works
.BMP or .DIB (device-independent bitmap) is a bitmapped In this paper we want to optimize the model of DOM which
graphics format used internally by the Microsoft Windows the Habibizad and cooperators proposed in [8]. Habibizad
and OS/2 graphics subsystem (GDI), and used commonly as and cooperators in [8] proposed:
a simple graphics file format on those platforms. This
format is in two parts, Part one, is the data, which indicate
whole image features including width, height, color palette 2.1 Data Oriented Modeling
etc. The second part consists a block of bytes that describes Habibizad and cooperators suppose that Fig1 is the image
the image, pixel by pixel. Pixels are stored starting in the for data oriented modeling. Fig. 2 shows ADBT of Fig. 1.
bottom left corner going from left to right and then row by ADBT is Average-Difference-Binary-Tree-of-Image which
row from the bottom to the top. Each pixel is described is stored in an array. This ADBT has three levels, numbered
using one or more bytes [9]. as 0, 1 and 2. The features of pixels are stored in the leaves.
Features of F are obtained by combining features of A and
C. In the same way, features of G are obtained from B and
(IJCNS) International Journal of Computer and Network Security, 69
Vol. 2, No. 3, March 2010

D. finally features of the entire image, E, are obtained by Fig. 6 shows segmented of Fig. 5. This segmented image is
combining features of F and G. By putting F and G together, created with histogram thresholding. Fig. 7 shows the
we can achieve smooth of original image as is shown in Fig. reverse N for image processing in this model. Fig. 8 shows
3 [8]. ADBT of Fig. 6. This ADBT has three levels, numbered as
0, 1 and 2. The features of regions are stored in the leaves.
Fig. 4 shows the array, which stores the ADBT of Fig. 2. Features of L0 are obtained by combining features of R0 and
R1. In the same way, features of L1 are obtained from R2
and R3. Finally features of the entire image, L2, are
obtained by combining features of L0 and L1. By putting L0
and L1 together, we can achieve smooth of original image
as is shown in Fig. 9. Fig. 10 shows the ADBT of Fig. 6 that
1
Figure 1. an I2 Image [8]. has been stored in an array.

In Fig.8 A and B and D have the similar intensity.

3.2 Edge detection with Sobel Algorithm


Edges characterize boundaries and are therefore a problem
of fundamental importance in image processing. Edges in
imag-

1
Figure 2. ADBT of an I2 image [8].

Figure 3. Smoothed of original image [8].


Figure 6. a segmented image.

Figure 4. ADBT of Fig. 2 stored in an array [8].

For ease of understanding, mentioned steps are described


top to down; but for enhancing speed of making ADBT, it
will be created down to top [8]. After that leaves will be
initialized by values of produced vector. Therefore, count of
leaves is equal to count of pixels of image. Each leaf is
corresponded to a pixel of original image and its A and D
are same as the pixel’s color. For each non-leaf node, A is
equal to average of its children’s A and D is equal to
difference of its children’s A [8].
Figure 7. the reverse N for image processing.

3. Our Works

3.1 Dynamic Data Oriented Modeling


This paper presents a dynamic data oriented model
(DDOM) of image. To illustrate the concept, Fig. 5 shows a
3 3 images, which have nine pixels, labeled A, B, C, D, E, F,
G, H and I respectively.

Figure 8. ADBT of Fig.8

Figure 5. An instance Image Figure 9. Smoothed of original image


70 (IJCNS) International Journal of Computer and Network Security,
Vol. 2, No. 3, March 2010

Figure 10. ADBT of Fig. 8 stored in an array

-es are areas with strong intensity contrasts – a jump in


intensity from one pixel to the next. Edge detecting an
image significantly reduces the amount of data and filters
out useless information, while preserving the important
structural properties in an image. There are many ways to
perform edge detection. However, the majority of different
methods may be grouped into two categories, gradient and
Laplacian. The gradient method detects the edges by looking
for the maximum and minimum in the first derivative of the
image. The Laplacian method searches for zero crossings in
the second derivative of the image to find edges [12].

The Sobel operator performs a 2-D spatial gradient


measurement on an image. The Sobel operator is shown in
Fig. 11.
Figure 12. Edge detection with Sobel algorithm
Note: G is Gradient and f is Image.
4.3 Image Segmentation Algorithm
According to the S. Arora and cooperator’s results in article
[13], segmentation algorithm is:

Following steps describe the proposed algorithm for image


segmentation:

1. Repeat step 2–6, times; where n is the


Figure 11. Sobel Operator. number of thresholds.
2. Range R = [a, b]; initially a = 0 and b = 255.
(1) 3. Find mean and standard deviation of all the
(2) pixels in R.
4. Sub-ranges’ boundaries and are calculated as
and ;
where and are free parameters.
5. Pixels with intensity values in the interval [ ]
(3)
and [ ] are assigned threshold values equal to
(4) the respective weighted means of their values.
6. a = T1 + 1 ; b = T2 − 1 .
7. Finally, repeat step 5 with and
with .
(5)
In Fig13 shown the result of image segmentation algorithm.
(6) Also in Fig14 the result of histogram Lena is shown.

In Fig. 12 Shown the result of Sobel algorithm:


4. Conclusion
In this paper, a dynamic data oriented model for image is
introduced which models image as a binary tree. Each node
of this tree represents the feature of image or sub image. By
using the presented model in this paper, tasks as measuring
similarity ratio, segmentation, clustering … can be done
with high desired precision and corresponding speed.
(IJCNS) International Journal of Computer and Network Security, 71
Vol. 2, No. 3, March 2010

segmentation through a fast statistical recursive


algorithm", pattern recognition letters 29 (2008) 119-
125, published by Elsevier.

Authors Profile
Asghar Shahrzad Khashandarag received the
B.S. degree in computer engineering from the
Payame Noor University Bonab Branch, Iran, in
2008, and the M.S. degree in computer
engineering from the Islamic Azad University
Tabriz Branch, Iran, in 2009, respectively.
From 2008, he works as a researcher with the Young Researchers
Club of Tabriz. He has published more than 10 papers in various
journals and conference proceeding. His research interests include
Figure 13. Results: Lena (a) Lena gray, (b) histogram, (c) 2 image processing, signal processing, wireless sensor network.
level thresholding, (d) 4 level, (e) 6 level and (f) 8 level.
Alireza Mousavi received his B.Sc. in computer
engineering, software Engineering, from
Allameh Mohaddes Nouri University,
Mazandaran, Iran, in 2008, and the M.S. degree
in computer engineering from the Islamic Azad
University Tabriz Branch, Iran, in 2010,
respectively. His research interests include image processing,
Residue Number systems, wireless sensor network.

Figure 14. Result of Histogram Ramin Aliabadian received the B.Sc. degree in
computer engineering from the Shomal
University, Amole, Iran, in 2008, and the M.S.
degree in computer engineering from the Islamic
Azad University Arak Branch, Iran, in 2010,
References respectively. His research interests include
[1] Zhiyong Wang, Dagan Feng, and Zheru Chi, “Region- image processing, computer architecture, computer networks.
Based Binary Tree Representation For Image
Classification”, IEEE International Conference on
Neural Networks & Signal Processing, Nanjing,
China, 2003.
[2] Xiaolin Wu, “Image Coding by Adaptive Tree-
Structured Segmentation”, IEEE Transactions on Davar Kheirandish Taleshmekaeil received
Information Theory, VOL. 38, NO. 6, 1992. the B.Sc. degree in Computer Hardware
[3] G.S.Seetharaman, B.Zavidovique, “Image Processing in engineering from the Allameh Mohaddes Nouri
a Tree of Peano Coded Images”, Computer Architecture University, Mazandaran, Iran, in 2008, and the
for Machine Perception, 1997.CAMP’97. Proceeding M.S. degree in computer engineering from
Fourth IEEE International Workshop on.
[4] M. Kunt, M. Benard, and R. Leonardi, “Recent results
the Islamic Azad University Tabriz Branch,
in highcompression image coding”, IEEE Transaction Tabriz, Iran, in 2010, respectively. His research interests include
Circuits Syst., vol. CAS- 34, no. 11, pp. 1306-1336, image processing, computer architecture, computer networks.
Nov 1987.
[5] R. Leonardi and M Kunt, “Adaptive split-and-merge for
image analysis and coding”, Proc. SPIE, vol. 594, 1985.
[6] G. J. Sullivan and R. L. Baker, “Efficient quadtree
coding of images and video”, in ICASSP Proc., May Ali Ranjide Rezai received the B.Sc. degree in
1991, pp. 2661-2664. computer engineering from the Shomal
[7] J. Vaisey and A. Gersho, “Image compression with University Amole, Iran, in 2008, and the M.S.
variable block size segmentation”, IEEE Trans. Signal
Processing, vol. SP-40, pp.2040- 2060, Aug. 1992. degree in computer engineering from the
Islamic Azad University Tabriz Branch, Iran, in
[8] A. Habibizad Navin, A. Sadighi, M. Naghian
Fesharaki, M. Mirnia, M. Teshnelab, and R. Keshmiri, 2010, respectively. His research interests
"Data Oriented Model of image : as a framework for include image processing, computer architecture.
image processing", World Academy of Science,
Engineering and Technology 34 2007
[9] http://en.wikipedia.org/wiki/Windows_bitmap
[10] http://en.wikipedia.org/wiki/GIF
[11] http://en.wikipedia.org/wiki/JPEG
[12] http://www.pages.drexel.edu/~weg22/edge.html
[13] S. Arora, J. Acharya , A. Verma, Prasanta K.
Panigrahi, " Multilevel thresholding for image

S-ar putea să vă placă și