Вы находитесь на странице: 1из 5

Middle-East Journal of Scientific Research 20 (3): 323-327, 2014

ISSN 1990-9233
IDOSI Publications, 2014
DOI: 10.5829/idosi.mejsr.2014.20.03.11457

Modified Algorithm Marker-Controlled Watershed Transform

for Image Segmentation Based on Curvelet Threshold

Med Ali Hamdi

Ecole Nationale dIngnieur de Tunis Laboratoires LTSIRS

Institut National des Sciences Appliques et de Technologie
BP 37 Belvdre 1002 Tunis Tunisie


Abstract: With the repaid advancements of computing technology, any use of the computer-based
technologies Increase in different scientific fields. The segmentations image is an important problem in
different fields of computer vision and image processing. The segmentation is a process of dividing image
according to its characteristic e.g., objects and color present in the images. There are different methods in
image segmentation. The focus of this study is a morphological operator called watershed segmentation.
The tool MATLAB is used in this study. New result of watershed segmentation entirely relay on the image
contrast. Image contrast may be degraded during image acquisition. So watershed algorithm can generate
under or over segmentation on badly contrast images. In order to reduce these deficiencies of watershed
algorithm a preprocessing step using Curvelet transform (CT) is performed on input images. Curvelet transform
is a strong approach, which improves the image contrast in the way image is degraded.
Key words: Image processing

Curvelet transform




watershed algorithm relies mostly on an estimation

of the gradients. The low-contrast [1] edges produce an
under segmentation and generate small magnitude
gradients, causing distinct regions to be erroneously
merged. Also over-segmentation and degraded watershed
produced by the noise.
So in order to reduce these deficiencies of watershed,
many preprocessing techniques are proposed by the
different researchers for example Moga presented an
efficient watershed algorithm based on connected
components, Jung and Scharcanski presented a
watershed segmentation using wavelets to denoise the
image. Bieniek and Hamarneh and Li presented a
watershed segmentation using appearance knowledge
and prior shape to improve their method etc. But most of
the methods proposed consider the over segmentation
problems and focus on the denoising of the image.
The under segmentation problem image and low
contrast are not yet addressed by most of the researchers.
The proposed technique focuses on the solution of
under segmentation problem of low contrast images
by applying preprocessing on the input image.

Segmentation is a basic problem in different fields,

in image processing for example image analysis, scene
analysis and pattern recognition. The segmentation is the
process of dividing images into regions where accord to
its characteristic e.g., objects and color present in the
images. These
have some meaningful
information about object and are sets of pixels.
The results of segmentation are in the form of images
that are more meaningful, easier to analyze and to
understand. In order to locate boundaries and objects
in images feature extraction of texture, object shape,
surface visualization and optical density, image
compression and image segmentation are used. Good
segmented result is very useful for the predication,
diagnoses and analysis. Watershed algorithm is a
morphological based method of the image segmentation.
The morphological gradient magnitude of an image is
considered as a topographic surface for the watershed
algorithm. The lines of watershed could be found by
different methods. The division of the image through

Corresponding Author: Med Ali Hamdi, Ecole Nationale dIngnieur de Tunis Laboratoires LTSIRS.


Middle-East J. Sci. Res., 20 (3): 323-327, 2014

The technique for preprocessing on the images is

Curvelet transform is a multiscale directional transform
which allows an almost optimal nonadaptive sparse
representation of objects with edges. It has generated
increasing interest in the community of applied
mathematics and signal processing over the past years. It
is an approach used to enhance the image contrast when
image is degraded.

Where L1, j, L2,


Calculate the arithmetic mean of pixels of

distorted image, arithmetic mean indicate to
central point of all values of pixels of distorted
images [7, 8].
Compute the Spatial Frequency Measure that
characterizes the distorted image.
Compute Difference Operator that detect edges of
distorted image and detect the slope of the gray
levels in the distorted image.
Then we compute threshold as the following:
Threshold = (difference operator of distorted image) *
(Spatial frequency measurement of distorted image) *
mean (distorted image)+(mean(distorted image) *
(number of coefficient of Curvelet), Where number of
coefficient of Curvelet equal number of iteration [9].
Third step we apply warping Fast Discrete Curvelet
Transform which can be calculated using the following
equation: [10]

The curved singularities should be approximated

with some coefficients and in a non-adaptive
representation named curvelets."
The curvelets remain coherent waveforms under the
action of the wave equation in a smooth medium

{ } is a wave packet frame on L R


based on second dyadic decomposition:

f ( x) = C
( x ) And C = f ( x )
the Frequency shells is 2k <
Sectors is

( x ) dx ;


C D , N ( j , l, k=

< 2k +1 ; and the Angular

) < 2 k / 2 finally the approximation rate

We choose n largest coefficients C in f = C

and f f 2 L2 < n 2 log ( n )2 .
And no frame can do better for jumps along C2
The wavelet expansion: f f 2 L2 < n 2 .

k n +k n
1 1 2 2
R1, j R2, J


The Watershed Transform: The watershed transform

[11] is a morphological based tool for image
segmentation. In grey scale the mathematical
morphology watershed transform for segmentation is
originally proposed by Digabel and Lantuejoul
in1977 and later improved by Li et Al in 2003. The
watershed transform can be classified as a region-based
segmentation approach.
The idea [12] of watershed can be view as a
landscape immersed in a lake; catchment basins will be
filled up with water starting at each local minimum.
Dams must be built where the water coming from different
catchment basins may be meeting in order to avoid the
merging of catchment basins.

The Adaptive Threshold Curvelet: The proposed method

[6] enhance gray scale images that has various types of
noises such as speckle, Gaussian, salt and pepper and
random noise. The method divides into the following
Firstly we compute norm of the Curvelet transform
where Curvelet normalized using the following equation:
C D ( j , l, k )
L1, j L2, j



Fourth step we apply computed threshold on

distorted image, then we apply inverse FDCT on
distorted image to move image from curvelet to spatial
domain. Finally we remove imagine part of image that
obtained from above step to obtain enhanced gray
scale image.

is optimal:

C D , N ( j, l , k ) =

are the side lengths of the

Secondly We Compute Threshold in Three Steps:

The Curvelet Transform: The Curvelet transform is a

transform and a
dimensional of the Wavelet transform which allows an
optimal non adaptive representation of edges designed
to represent images at different scales and different
angles. [2, 3] Curvelet has two mathematical properties
[4, 5]:

A curvelet frame [6]



Middle-East J. Sci. Res., 20 (3): 323-327, 2014

watershed algorithm. The process of rain falling is

described in Figure 2.
Marker-Controlled Watershed Segmentation:Separating
touching objects in image is one of more difficult image
processing operations. The watershed transform is often
applied to this problem. The watershed transform finds
catchment basins CB and watershed ridge lines in an
image by treating it as a surface where light pixels are high
and dark pixels are low. [14, 15] The segmentation using
the watershed transform works better if we can identify,
or mark, the foreground objects and the background
locations, so Marker controlled watershed segmentation
follows this procedure [16], First we compute a
segmentation function, then we compute foreground
markers, after we compute background markers and we
modify the segmentation function so that it only has
minima at the foreground and background marker
locations. Finally we compute the watershed transform
of the modified segmentation function.

Fig. 1: Illustration of immersion process of watershed

transforms. (CB: Catchment basins)

Proposed Technique: The proposed image segmentation

algorithm here is able to segment the images with
minimum drawbacks of under segmentation and over
segmentation. The steps of the proposed system are
shown in Figure 3.
results of the proposed techniques are
described in this section, we used MIAS database for
As the results shows that the segmented images with
the pre processing using adaptive Threshold Curvelet
has less chances of under or over segmentation.
The difference between the segmentation using adaptive
Threshold Curvelet and without using adaptive Threshold
Curvelet as a preprocessing step is given. This shows the
improvement in the segmentation results.

Fig. 2: Illustrations of flooding (process of watershed

The water shed lines are defined by the catchment
basins divided by the dam at the highest level where the
water can reach in the landscape [13]. As a result,
watershed lines can separate individual catchment
basins in the landscape. The idea is described in Figure 1
which describes the flooding or rain falling process of

Fig. 3: Proposed process for image segmentation


Middle-East J. Sci. Res., 20 (3): 323-327, 2014

Fig. 4: Original image and segmentation results


The system will examined only one image at a

time. This system can be very helpful for the
segmentation of the images which are used in
different fields of life. And the image analysis
process can be facilitated by this system. The research
content of this system was segmentation and image

The goal of image segmentation process is to

identify the segments of the image according to the
image characteristic e.g. objects shape, image color, etc.
The simplified working of the image segmentation system
is stated here. The most important step is the image
acquisition. Any deficiency during the image acquisition
can cause many problems in the result. The image used in
this process is taken from MIAS database. The input
images are of low contrast. This segmentation process
deals with the problem caused by these low contrast
images by applying a preprocessing step using adaptive
Threshold Curvelet. This step enhances the contrast of
the input image so that the gradient of the image is strong
enough to properly segment the image by using the
watershed. After preprocessing step the gradient of the
image is finding by converting the input image to grey
scale. And this gradient of image is used as the input the
image. The results show the improvement in the
segmentation results using adaptive Threshold Curvelet.




Debeir, O., I. Adanja, N. Warzee, P. Van Ham and

C. Decaestecker, 2008. Phase contrast image
segmentation by
weak water shedtransform
assembly Biomedical Imaging: From Nano to Macro,
2008. ISBI 2008. 5th IEEE International Symposium
Digital Object Identifier:10.1109/ISBI.2008.4541098
Publication, 724-727 IEEE conferences.
Donoho, D.L.
Duncan, 1999.

Middle-East J. Sci. Res., 20 (3): 323-327, 2014








Starck, J.L., E. Candes and D.L. Donoho, 2002.

The Curvelet Transform for Image Denoising. IEEE
Transactions on Image Processing. 11(6): 670-684.
Jean-Luc Starck, Fionn Murtagh, Jalal M. Fadili,
sparse image and signal processing Wavelets,
Curvelets, Morphological Diversity cambridge
university press Cambridge, New York, Melbourne,
Madrid, Cape Town, Singapore, Sao Paulo, Delhi,
Dubai, Tokyo Cambridge University Press 32
Avenue of the Americas, New York, NY 10013-2473,
USA www.cambridge.org Information on this title:
Starck, J.L., E. Candes, F. Murtagh and D.L. Donoho,
2003. Gray and Color Image Contrast Enhancement
by the Curvelet Transform. IEEE Transaction on
Image Processing, 12(6): 706-717.
Gang Tang and Ma. Jianwei, 2011. Application of
Total-Variation-Based Curvelet
Shrinkage for
Three-Dimensional Seismic Data Denoising
Geoscience and Remote Sensing Letters, IEEE
Identifier:10.1109/LGRS.2010.2052345 Publication
Year: 2011, Page(s): 103-107 IEEE JOURNALS.
Jianwei, Ma and Gerlind Plonka, A Review of
Curvelets and Recent Applications, School of
Aerospace, Tsinghua University, Beijing 100084,
China Centre de Geosciences, Ecole des Mines de
Paris, 77305 Fontainebleau Cedex, France Faculty
of Mathematics, University of Duisburg-Essen, 47048
Duisburg, Germany.
Jiulong Zhang and Wang Yinghui, 2010.
A comparative study of wavelet andCurvelet
transform for face recognition Image and Signal
Processing (CISP), 2010 3rd International Congress
Volume: 4 Digital Object dentifier:10.1109/
CISP.2010.5647882 Publication, 1718 1722 IEEE
Zhang, Zhiyu, Zhang Xiaodan and Zhang Jiulong,
2009. SAR Image Processing Based on Fast discrete
CurveletTransform Information Technology and
Applications, 2009. IFITA'09. International Forum 3
Digital Object Identifier:10.1109/IFITA.2009.124
Publication,: 28-31 IEEE conferences.


10. Cand`es, E.L., Demanet, D. Donoho and L. Ying, 2006.

Fast Discrete Curvelet Transforms, http://
11. Yuqian Zhao, Liu. Jianxin, Li. Huifen and Li. Guiyuan,
2008. Improved water shedalgorithm for dowels
image segmentation Intelligent Control and
Automation, 2008. WCICA 2008. 7th World Congress
WCICA.2008.4594115 Publication, 7644 - 7648 IEEE
12. Cheng Sheng-dong, Li Peng, Li Zhan-bin and
Lu Ke-xin, 2011. Study on topographic characters of
soil erosion spatial distributionin inwatershedon
Loess Plateau Electric Technology and Civil
Engineering ICETCE, 2011 International Conference
ICETCE.2011.5776039Publication, 5563-5566 IEEE
13. Wenli, Yang, Zeng. Zhiyuan and Zhang. Sizhe, 2010.
Application of Combining Watershed and Fast
Clustering Method in Image Segmentation Computer
Modeling and Simulation, 2010. ICCMS '10. Second
International Conference 3 Digital Object
Page(s): 170-174 IEEE CONFERENCES.
14. Heijmans, H.J.A.M., 1997. Composing Morphological
Filters, IEEE, 6(5): 713-723.
15. Ghassan Hamarneh and Li. Xiaoxing, 2009.
Watershed segmentation using prior shape and
knowledge Image and
Computing, Canadian Robotic Vision 2005 and 2006,
27(1-2): 59-68.
16. Shuang Wang, Ma. Xiuli, Zhang. Xiangrong and
Jiao. Licheng, 2007. Watershed-based textural image
segmentation Intelligent Signal Processing and
Communication Systems, 2007. ISPACS 2007.
International Symposium on Digital Object
312-315, IEEE conferences.