Content uploaded by Krishnan Nallaperumal
Author content
All content in this area was uploaded by Krishnan Nallaperumal
Content may be subject to copyright.
Abstract — A novel multiscale morphological watershed
segmentation algorithm for the segmentation of color images is
proposed in this paper. The conventional watershed segmentation
algorithm is not suitable for color image directly, since it is hard to
decide a suitable color gradient measurement for color image. The
proposed algorithm uses HSV color space to define color contrast
gradient which is multiplied with multiscale morphological gradient
of the intensity image. The marker extractions of this composite
color gradient image are obtained by a thresholding technique to
avoid over segmentaion vivid in conventional watershed algorithm
and given as input to the watershed algorithm which uses Hill-
Climbing approach to identify and label the neighbourhood pixels
in the marker image. Experimental results show that the proposed
algorithm avoids the over segmentation problem and produce the
segmented result which is good for human perception.
I. INTRODUCTION
MAGE Image segmentation, a very important image analysis
task in digital image processing causes the fragmentation of
image into regions of homogeneity. It is used as a
preprocessing stage for numerous computer vision
applications such as pattern recognition, image retrieval and
smart surveillance. But it is still a difficult task to construct a
general method which tends to produce perfect meaningful
segments for the variant natural images.
The segmentation techniques for intensity images like
feature thresholding, contour based techniques, region based
techniques, clustering, template matching and watershed
segmentation etc have their own advantages and limitations in
terms of applicability, suitability, performance and
computational cost. It is difficult that all these qualities are
simultaneously felt by a single segmentation algorithm. Most
of the above approaches have been extended to color images,
but the related literatures are still short. In recent researches,
for natural color image segmentation, the most popular
algorithms usually are region-based algorithms, especially
combining with clustering method. These methods employ the
region-based partitioning technique that groups homogenous
Nallaperumal Krishnan is with the Centre for Information Technology and
Engineering, Manonmaniam Sundaranar University, Tirunelveli, India (phone:
+91-462-2323650;fax:+91-462-2334363;e-mail:krishnan17563@yahoo.com).
K.Krishnaveni is with the Department of Computer Science, Sri
S.Ramasamy Naidu Memorial College, Sattur, India (e-
mail:kkveni_srnmc@yahoo.co.in )..
pixels as particular regions and guarantees that the contours
are continuous with fixed width wide. For intensity image, the
region-based partitioning technique has three popular
methods: split-and-merge, region-growing, and watersheds.
But these methods cannot be applied to color images directly.
The multiscale analysis and hierarchical structure are the most
popular extended techniques for conventional algorithms.
They are successful methods, but usually their accurate effect
cost much higher computation [1] [3] [4].
Morphological segmentation algorithms are also more
common in literature [10-12]. Mathematical morphology, a set
theoretic, shape oriented approach treats the image as a set and
the kernel of operation commonly known as structuring
element (SE) as another set. Different standard morphological
operations like dilation, erosion, opening, closing are basically
set theoretic operations between these two sets. The shape and
the size of the SE play an important role in detecting and
extracting features of shape and size of the objects in the
image, its different sizes can be accounted by the ‘scale’
attribute of the structuring element. Thus morphological
operations with such scalable SEs can be used in multiscale
image segmentation.
Watershed segmentation, a very prominent segmentation
scheme has many advantages for image segmentation, such
that it ensures the closed region boundaries and gives solid
results. But it poses the serious over-segmentation problem.
This problem is due to the noise particles which lead to
roughness of morphological gradient. A suitable
preprocessing technique will give proper gradient definition.
If the edge between particular regions could be strengthened
and the protuberance over the plains could be smoothened, the
gradient image would be more obvious to reduce the over-
segmentation problem and a better result is expectable [9]
[12].
In this paper, we propose a novel segmentation algorithm
that works on the marker extractions of the composite color
gradient image obtained by multiplication of color contrast
gradient and multiscale intensity gradient, which are
segmented by watershed algorithm using Hill- Climbing
technique. The proposed algorithm can work well on color
images and reduce the limitations of many other algorithms.
The paper is organized in V sections. Section II discusses
the Color model and Watershed transform. Section III
describes the features and the scheme of proposed
A Multiscale Morphological Watershed
Segmentation using Color Composite Gradient
and Marker Extraction
Krishnan Nallaperumal1, Senior Member, IEEE, Krishnaveni. K2, Member, IEEE
I
INTERNATIONAL JOURNAL OF IMAGING SCIENCE AND ENGINEERING (IJISE)
IJISE,GA,USA,ISSN:1934-9955,VOL.2,NO.2, APRIL 2008
195
segmentation algorithm. The Experimental analysis and
simulation results are presented in Section IV. Conclusions
are finally made in Section V.
II. COLOR MODEL AND WATERSHED TRANSFORM
A. Color Model
For segmenting color images, the first step is to choose a
perfect color model. Color is a very useful feature in the
analysis of image content for segmentation or retrieval.The
distinguishable characteristics of color are hue, saturation and
brightness and the mixture of hue and saturation is defined as
chromaticity. The conventional representation of color image
is by red-green-blue (RGB) color model. Since the RGB
components are highly correlated, it is not convenient to make
the chromatic information useful directly.
HSV model is a nonlinear transformation of RGB color
space and is commonly used in computer applications. The
Hue is what people normally think of as color. The Saturation
is the amount of gray, white or black that is mixed to the
color. Zero saturation indicates no hue just gray scale. The
Value component of HSV indicates measure of its brightness.
The HSV model is shown in fig.1. A color in HSV space is
specified by stating a hue angle, the saturation level and the
value level. A hue angle of zero represents red color.
Fig. 1 HSV color space as a conical object
There are seven color contrast concepts: hue, light-dark,
cold-warm, complementary, simultaneous, saturation and
extension [1]. The contrast of hue relates to mixture lineage
on the color wheel; where primaries such as red, yellow and
blue stand as strongest contrast. The light dark contrast deals
with the intensity dimension of color perception. The two
extremes are black and white with grays and chromatic hues
lying in between. The contrast of saturation concerns with
perceptual color purity between vivid saturated and pale
distributed colors. Contrast of saturation is a relative measure.
A color may appear vivid beside a diluted tone, but dull beside
a more brilliant tone.
Fig. 2 shows a HSV color wheel which consists of an outer
halo and an inner roundlet with different saturation and
brightness value. As shown in this color wheel, colors are
perceived mainly by hue, and also influenced by the saturation
and brightness. The higher the brightness and saturation arc,
the more colors are distinguishable [1] [2].
Fig. 2 The HSV color wheel, (a) the brightness and
saturation are different between inner and outer circles.
(b) Hue of 36 levels
B. Watershed Transform
The watershed transform is a popular segmentation method
in the field of mathematical morphology. It is considered to be
a topographic region growing method. The intuitive idea
underlying this method comes from geography: it is that of a
landscape or topographic relief which is flooded by water,
watersheds being the divide lines of the domains of attraction
of rain falling over the region. An alternative approach is to
imagine the landscape being immersed in a lake, with holes
pierced in local minima. Basins (also called ‘catchment
basins’) will fill up with water starting at these local minima,
and, at points where water coming from different basins
would meet, dams are built. When the water level has reached
the highest peak in the landscape, the process is stopped. As a
result, the landscape is partitioned into regions or basins
separated by dams, called watershed lines or simply
watersheds. Fig. 3 illustrates the concept of watershed.
Fig. 3 Minima, catchment basins, and watersheds
The watershed transform has been widely used in many
fields of image processing, including medical image
segmentation, due to some advantages given below.
• It is a simple and intuitive method.
• The watershed lines always correspond to the most
significant edges between the markers. So this technique is
not affected by lower-contrast edges, due to noise, that
could produce local minima and, thus, erroneous results, in
energy minimization methods.
• Even if there are no strong edges between the markers, the
watershed transform always detects a contour in the area.
KRISHNAN NALLAPERUMAL, KRISHNAVENI : A MULTISCALE MORPHOLOGICAL WATERSHED SEGMENTATION ... MARKER EXTRACTION
IJISE,GA,USA,ISSN:1934-9955,VOL.2,NO.2, APRIL 2008
196
This contour will be located on the pixels with higher
contrast.
Some important drawbacks also exist,
¾ Over-segmentation
When the watershed transform infers catchment basins from
the gradient of the image, the result of the watershed
transform contains a myriad of small regions, which makes
this result hardly useful. The use of a marker image to reduce
the number of minima of the image and, thus, the number of
regions, is the most commonly used solution. Also interesting
is the utilization of a scale space approach to select the
interesting regions, using different filters.
Fig .4(a) Original image (b) conventional watershed
segmentation.
¾ Sensitivity to noise
Local variations of the image can change dramatically the
results. This effect is worsened by the use of effective
preprocessing technique.
¾ Poor detection of significant areas with low contrast
boundaries
If the signal to noise ratio is not high enough at the contour
of interest, the watershed transform will be unable to detect it
accurately. Furthermore, the watershed transform naturally
detects the contours with higher value between markers,
which are not always the contours of interest.
¾ Poor detection of thin structures
When the watershed transform is applied on the gradient
image, the smoothing associated with gradient estimation,
together with usual approach of storing gradient values only at
the image pixel positions rather than with sub-pixel accuracy,
make it difficult to detect thin catchment basin areas [5] [9]
[10].
To overcome these problems, we propose a new composite
color gradient technique and the resultant gradient image is
marker extracted, to reduce the number of local minima. The
proposed watershed segmentation makes use of Hill-Climbing
Technique for efficient segmentation results.
III. THE PROPOSED MULTISCALE MORPHOLOGICAL
WATERSHED SEGMENTATION USING COLOR COMPOSITE
GRADIENT AND MARKER EXTRACTION
The proposed segmentation algorithm is a robust technique
that can be applied on color images. The segmentation
algorithm is more computationally efficient, its edge-
preserving, noise removing, scale-calibrating, shape
maintaining features are remarkable than many top-ranking
color image segmentation algorithms. The noisy image is
preprocessed and smoothened by using morphological
operators. The preprocessed image is transformed to HSV
color space representation in order to analyze and establish a
color contrast gradient. The multiscale morphological gradient
of the intensity channel of original image is obtained and
multiplied with color contrast gradient to give composite color
gradient image. Marker extractions of this image is obtained
and given as input to the watershed algorithm which uses a
Hill-Climbing approach to identify and label the
neighborhood pixels. The proposed segmentation algorithm is
schematically illustrated in the block diagram shown in fig. 5.
Fig. 5 Block Diagram of the proposed algorithm
A. Preprocessing
A segmentation algorithm can give better output without
over-segmentation on a preprocessed noise free image. Images
can get corrupted by additive impulse noise in the acquisition
or transmission stages. Such noisy images can be
preprocessed and smoothened by a set of morphological
operators [10].
B. Converting RGB to HSV
The proposed algorithm aims to segment the given image
with visually distinct colors, so a human perception based
Composite Color Gradient
Input RGB Image
Color Contrast
gradient
Multiscale Morphological
Gradient
Marker Extraction
Watershed segmentation
Convert RGB to
HSV
Morphological
smoothening
Quantize HSV
values
Construct Intensity
image
Segmented Image
INTERNATIONAL JOURNAL OF IMAGING SCIENCE AND ENGINEERING (IJISE)
IJISE,GA,USA,ISSN:1934-9955,VOL.2,NO.2, APRIL 2008
197
color model is preferred. The representation of color images
by Red-Green-Blue color model is the conventional method.
But the three components of RGB color model are highly
correlated, so that the chromatic information is not suitable for
using directly. Therefore, the HSV color space is adopted
because it is convenient to convert from RGB and also
intimately related to human perception [2]. The conversion
from RGB to HSV is given below.
Max=max(R, G, B); Min=min(R, G, B);
Value= max(R, G, B);
If (Max==0) then
Saturation=0;
Else
Saturation= (Max-Min)/Max;
If (Max==Min | Hue is undefined (achromatic color)
Otherwise:
If (Max==R && G>B)
Hue= (60 * (G-B)/(Max-Min)
Else if (Max==R && G<B)
Hue=360+ (60*(G-B)/(Max-Min)
Else if (G ==Max)
Hue=60*(2.0+ (B-R)/ (Max-Min)
Else Hue=60*(4.0+(R-G)/ (Max-Min))
C. Quantization in HSV space
For the purpose of human perception and simplifying the
image, the (H, S, V) values are uniformly quantized into
(37, 5, 9) levels respectively. The five levels of saturation
represent five grade of chroma: achromatic, nearly
achromatic, low achromatic, middling chromatic, and highly
chromatic. The range of quantized saturation is from 0 to 4. In
the achromatic level, the saturation value is equal to zero and
there is no perceived color. In the nearly achromatic level, the
color is perceived as an achromatic color even if a color has a
hue component. In the above situation, the perceptual contrast
depended on the brightness. In low or middling chromatic
level, the number of perceptible colors is still slightly less than
in highly chromatic level. In these situations, the perceptual
contrast depended on hue mainly.
In general human can distinguish some colors: red, green,
blue, yellow and mixed colors such as orange. The named
colors will be in non-uniform quantized position of the hue
plane and it is difficult to measure the distance of two colors.
To reduce this variation, quantization process is intended for
calculating color contrast. The original hue plane is divided
into 37 levels. The levels of value are quantized to 9 levels
and saturation into 5 levels based on the following set of
equations. Finally, the (qH, qS, qV) represents the quantized
(H, S, V)[1].
()
()
()
H [0, 360]; qH [0, 36]; qH is integer.
0 if H=0, No Hue
qH =
H-5 /10+ 1) mod (36)
S [0,1] ; qS [0,4] ; qS is integer
qS = S * 100 / 25
V [0,255] ; qV [0, 8] ; qV is integer.
qV = V / 32
∈∈
⎧
⎪
⎨
⎪
⎩
∈∈
∈∈
D. Color Contrast Gradient
The difference of Hue, Saturation and value are constructed
by considering the eight neighbors of each pixel in the
quantized HSV image. Since the circular relationship of hue,
the maximal difference of hue between two pixels is defined
as 18, and the distance between level 36 and level 1 is equal to
1. Consequently the maximal distance between two pixels is
18+4+8=30[1].
Since the perceptible colors vary with different chromatic
condition, and two colors are perceptibly different when the
difference of their quantized hue level is more than three in
the highly chromatic situation.
Beside, according to color relationship introduced above,
saturation represents the degree how much white element is
mixed to a pure color. In human perception, saturation often
reflects the intensity of lightness, and the brightness is more
perceptible than hue in the achromatic and nearly achromatic
situation. Therefore saturation plays a critical role in our color
contrast measurement; it is a criterion to determine how much
the difference of quantized hue (qH) to settle one degree of
hue and brightness contrast when in the different chromatic
condition.
In order to emphasize the importance of hue, the color
contrast matrix, GRAD, is defined as follows:
GRAD = Crt_Vx # wtv + 2 Crt_Hx # wth (1)
where, Crt_V and Crt_H are the difference matrices of
brightness and hue of pixel (i,j) related to its eight neighbors,
wtv and wth are the related weight matrix depended on
saturation level.
Consequently, the color contrast gradient (CCG) of each
pixel is whose maximal GRAD, as follow:
CCGi, j = MAX (GRAD) (2)
Finally the gradients are normalized to be in the range from
0 to 10 and a normalized color contrast gradient (NCCG) is
obtained.
NCCGi,j = 10[CCGi,j/ MAX(CCGi,j )] (3)
D. Multiscale Morphological Gradient
The intensity channel of original image is smoothened to
keep the interior of the objects and to preserve the boundary
of the objects. The objects to be smoothened are subjected to a
group of morphological operators. The basic morphological
operators involved in this phase are listed below.
In the morphological analysis of gray- scale images, a 2-D
image is defined as a subset of the 2-D Euclidean space RxR
or its digitized equivalent ZxZ. In this paper, we deal only
with intensity image that is defined as subsets of ZxZ.
The two most fundamental morphological operations are
dilation and erosion. Dilation of the image, f by the 4 or 8
KRISHNAN NALLAPERUMAL, KRISHNAVENI : A MULTISCALE MORPHOLOGICAL WATERSHED SEGMENTATION ... MARKER EXTRACTION
IJISE,GA,USA,ISSN:1934-9955,VOL.2,NO.2, APRIL 2008
198
connected structuring element (SE), B expands the image
while the erosion of f by B shrinks the image. They are
defined respectively with f and B in the set ZxZ as
f ⊕ B = {X/ (B)x ∩ f ≠ φ} (4)
f θ B = {x/ (B) x ⊆ f ≠ φ} (5)
Opening of the binary image f by the 4 or 8 connected
structuring element B denoted as f º B, is defined as
f º B = ( f θB) ⊕ B (6)
Closing of the binary image f by the 4 or 8 connected
structuring element B denoted as f • B, is defined as
f •B = ( f ⊕ B) θ B (7)
The local gray-level variation in the image can very well be
given by the morphological gradient. A gradient helps
detecting ramp edges and avoids thickening and merging of
edges providing edge-enhancements. The gradient image, G
(f) is morphologically obtained by subtracting the eroded
image, ε(f) from its dilated version, ∂(f). A multiscale
gradient, MG (f) is the average of morphological gradients
taken for different scales of the structure element, Bi.
where Bi is a SE of size (2i+1) x (2i+1) [13-14].
E. Composite Color Gradient Image
Composite Color gradient is obtained by multiplying
both color contrast image NCCG and multiscale
morphological gradient image MG (f).
F. Marker Extraction
The Watershed segmentation algorithm applied directly to
the composite color gradient image can cause over-
segmentation due to serious noise patches or other image
irregularities. The concept of Markers can be used to solve
this over-segmentation problem whose goal is to detect the
presence of homogeneous regions from the image by a
thresholding technique. They spatially locate object and
background, ensures to keep up the interior of the object as a
whole. The Markers are connected components belonging to
an image; internal markers are inside each of the objects of
interest and external markers are contained within the back-
ground. Here, the composite color gradient image is
thresholded to extract the markers. The resulting marker
image M(f) is a binary image such that a pixel is a marker (to
be black) if it belongs to a homogeneous region, a pixel will
be white if it does not belongs to homogeneous regions. Thus
the marker image contains a set of black pixels, i.e., Markers,
marking the core regions, and a set of white pixels remaining
unassigned to any regions. The improper selection of
threshold value will lead to over segmentation or under
segmentation [13 -15].
H .Watershed Transform
The fast implementation of immersion-based watershed
algorithm has been introduced by Vincent and Soille [9]. The
authors simulated a flooding process, in which the water
comes up out of the ground and floods the catchment basins
without predetermining the regional minima. Alternatively, an
ordered queue based watershed algorithm has been proposed
by Meyer [16]. This algorithm determines the regional minima
and starting from these minima, the recursive label
propagation is performed using an ordered queue. It yields a
complete tessellation of the image into its homogeneous
regions without producing any watershed lines. Finally,
several shortest-path algorithms for the watershed
transformation with respect to topographical distance can be
found in literature [5]. Here we proposed a fast watershed
transform based on Hill Climbing technique [6].
Since marker extracted composite color gradient image is
given as input to the Hill Climbing technique, number of local
minima is reduced and better segmentation result is obtained.
The complexity of the algorithm has been reduced by doing
away with multiplication normally required to form a lower
complete image in an intermediate step of the overall
segmentation process. Its moderate complexity makes it
amenable to dedicated hardware implementation.
IV. EXPERIMENTAL RESULTS
The proposed segmentation algorithm is tested on a wide
variety of natural color images for subjective analysis. The
algorithm gives optimal results even when the images are
highly corrupted with salt & pepper impulse noise. The fig
shows three different gradient images and segmentation
results.
In comparison to color contrast and multiscale
morphological gradient, the edges in the color composite
gradient are strengthened and the inner regions are
smoothened. The segmented results of this gradient image
have over segmentation problem. When we apply watershed
segmentation algorithm to marker extracted image many
smaller regions are removed and over segmentation problem
has been reduced.
V. CONCLUSION
A novel multiscale morphological watershed-based color
image segmentation algorithm is presented in this thesis. In
order to perform the watershed segmentation in color images
and diminish the over-segmentation problem, a combined
color gradient is proposed. The combined color gradient is the
product of the color contrast gradient and the multiscale
morphological intensity gradient. In our experiments, the
results produced by proposed algorithm are acceptable and
better than conventional watershed based methods. Our
algorithm can be applied to other applications, such as content
based image retrieval. The multi-gradient contrast can also be
applied to other segmentation algorithms.
The proposed algorithm can be improved by adapting
efficient low-pass filters for preprocessing and fuzzy optimal
threshold selection for marker extraction.
INTERNATIONAL JOURNAL OF IMAGING SCIENCE AND ENGINEERING (IJISE)
IJISE,GA,USA,ISSN:1934-9955,VOL.2,NO.2, APRIL 2008
199
Fig .6. Images segmented by Proposed Algorithm (a) Original Image (b) Composite color gradient
(c) Proposed segmentation Algorithm.
(a) (b) (c)
VI. REFERENCES
[1] Chao-yu chi and shen-chaun tai “perceptual color contrast based
watershed for color image segmentation”, IEEE, Oct 06.
[2] B.S. Manjunath, Jens-Rainer Ohm, Vinod V. Vasudevan, and Akio
Yamada, " Color and Texture Descriptors," IEEE Trans. Circuits
Syst.Video Technol., vol. 11, no.6, pp. 703-715, Jun. 2001
[3] Iris Vanhamel, loannis Partikakis, and Hichem Sahli, "Multiscale
Gradient watersheds of Color Images," IEEE Trans. Image
Processing,vol. 12, no. 6, pp. 617-626, Dec. 2003.
[4] John M. Gauch, "Image Segmentation and Analysis via Multiscale
Gradient Watershed Hierarchies," IEEE Trans. Image Processing, vol.8,
no.1, pp. 69-79, Jan. 1999.
[5] Roerdink, J.Meijster, A. “The watershed transform: Definitions,
algorithms and parallelization strategies”, Fundamental Informaticae
41(2000) 187-228.
[6] C.Rambabu, T.S.Rathore, I.Chakrabarti, “A new watershed algorithm
based on Hillclimbing Technique for Image segmentation”, IEEE, 2003.
[7] Jose A. Lay and Ling Guan, "Retrieval for Color Artistry concepts,"
IEEE Trans. Image Processing, vol. 13, no. 3, pp. 326-339, Mar. 2004.
[8] Kostas Haris, Serafim N. Efstratiadis, Nicos Maglaveras, and Aggelos
K. Katsaggelos, "Hybrid Image Segmentation Using Watersheds and
Fast Region Merging," IEEE Trans. Image Processing, vol. 7, no. 12, pp.
1684-1699, Dec. 1998.
[9] Luc Vincent and Pierre Soille, “Watersheds in digital spaces: An efficient
algorithm based on immersion simulations," IEEE Trans.Pattern and
Machine Intelligence, vol. 13, pp. 583-598, 1991.
[10] Rafael C. Gonzalez and Richard E. Woods, Digital Image Processing,
Prentice Hall, New Jersey, USA, pp. 567-635, 2001.
[11] Shao-Yi Chien, Yu-Wen Huang, and Liang-Gee Chen, "Predictive
Watershed: A Fast Watershed Algorithm for Video Segmentation," IEEE
Trans. Circuits Syst. Video Technol., vol. 13, no. 5, pp. 453-461, May.
2003.
[12] Yi-Chen Wu, "Watershed Segmentation of Color Image Based on Human
Perception," M.S. Thesis, Department of Computer Science and
Information Engineering, National Chung Cheng University, Jul. 2000.
[13] “An efficient Multi-scale Morphological Watershed Segmentation using
Gradient and Marker Extraction”, proc. of IEEE INDICON 2006,
New Delhi, India, September 2006.
[14] Krishnan Nallaperumal , K.Krishnaveni et al, “Fuzzy Optimal
Thresholded Multi-scale Morphological Segmentation of Digital
Images”, proc. of IEEE WOCN 2006, Bangalore, India, April 2006
[15] Krishnan Nallaperumal , K..Krishnaveni et al, “A Novel Multiscale
Morphological Watershed Segmentation Algorithm”. proc. of
ICACCC, Madurai, India, Feb.,2007.
[16] S. Beucher and F. Meyer, .The morphological approach to segmentation:
The watershed transformation,. In Mathematical Morphology in Image
Processing. New York : Marcel Dekker Inc., 1993, pp. 433.481
Nallaperumal Krishnan received M.Sc.
degree in Mathematics from Madurai
Kamaraj University, Madurai, India in 1985,
M.Tech degree in Computer and
Information Sciences from Cochin
University of Science and Technology,
Kochi, India in 1988 and Ph.D. degree in
Computer Science & Engineering from
Manonmaniam Sundaranar University,
Tirunelveli. Currently, he is the Professor
and Head of Center for Information
Technology and Engineering of Manonmaniam Sundaranar
University. His research interests include Signal and Image
Processing, Remote Sensing, Visual Perception, and mathematical
morphology fuzzy logic and pattern recognition. He has authored
three books, edited 18 volumes and published 25 scientific papers in
Journals. He is a Senior Member of the IEEE and chair of IEEE
Madras Section Signal Processing/Computational Intelligence /
Computer Joint Societies Chapter.
K.Krishnaveni received her B.E. degree
in Computer Science and Engineering
from Bharathiar University, Coimbatore,
India in 1990 and M.Tech degree in
Computer and Information Technology
from Center for Information Technology
and Engineering of Manonmaniam
Sundaranar University, Tirunelveli, India
in 2004 and is now pursuing her Ph.D in
the same University. Currently she is a
Selection-Grade Lecturer in Department of
Computer Science, Sri.S.Ramasamy Naidu Memorial College,
Sattur,India. Her research interests are Image Processing,
Mathematical Morphology and Fuzzy Logic. She is a member of the
IEEE.
KRISHNAN NALLAPERUMAL, KRISHNAVENI : A MULTISCALE MORPHOLOGICAL WATERSHED SEGMENTATION ... MARKER EXTRACTION
IJISE,GA,USA,ISSN:1934-9955,VOL.2,NO.2, APRIL 2008
200
