A version of watershed algorithm for color image segmentation

A version of watershed algorithm
for color image segmentation
Md. Habibur Rahman (11-94853-2)
Master’s Thesis Presentation and Defense
Thesis Committee :
American International University-Bangladesh
June, 2013
1
Prof. Dr. Md. Rafiqul Islam (Advisor)
Dr. Md. Saiful Azad (External)
Dr. Dip Nandi (Head of Graduate Program)

 Problem Definition
 Thesis Contributions
 Introduction
 Proposed Watershed Algorithm
 Image Quality Assessment (IQA) Metrics
 Results Analysis
 Conclusions
 List of publication
 References
Outline
2

 Over-segmentation problem in the existing
watershed algorithm
 Sensitive to noise
 High computational complexity
 Performance varies in different classes of
images
Problem Definition
3

 An adaptive masking and a thresholding
mechanism over each color channel before
combining the segmentation from each
channel into the final one
 Overcome over-segmentation problem
 Computationally inexpensive
 Perform well in case of noisy image
 Perform better with respect to five IQA
metrics in 20 different classes of images
Thesis Contributions
4

 What is digital image?
 Digital image processing
 How image is stored?
 Image Segmentation
 Why Image Segmentation?
 Color Image Segmentation Algorithms
Introduction
5

What is a digital image?
• A numeric
representation of a
two-dimensional
image as a finite set of
digital values
• Pixel values usually
represent intensity
levels or gray levels,
colors, heights, and
opacities [11].
611. R C Gonzalez and R E Woods, Digital Image Processing, 3rd Edition, Pearson, pp. 51

 An image can be defined as a two-
dimensional function, p (x, y)
 Where x and y are spatial (plane)
coordinated
 The amplitude of p at any pair of coordinates
(x, y) is called the intensity or gray level of
the image at that point
Digital Image Processing
7

How image is stored?
• In image, P (0, 0)
represents the top left
corner pixel
• P (X−1, 0) represents
the bottom left corner
pixel of the image
• In digital image, pixels
contain color value and
each pixel uses 8 bits or
1 Byte or 256 values [13]
813. H. Vankayalapati, "Evaluation of wavelet based linear subspace techniques for face recognition," Klagenfurt, 2008

 It is a process to divide the digital image into
homogeneous and different meaningful
regions
 The main goal of image segmentation is to
cluster of pixels in the relevant regions
 It is used to recognize similar regions and
grouping the similar visual objects
 Property like grey level, color, intensity,
texture, shape, depth or motion from the
digital image
Image Segmentation
9

 We do image segmentation to separate
homogeneous area
 It requires everywhere for precise
segmentation if we want to analyze what
inside the image.
 It is separate objects and analyze each
object individually to check what it is.
Why Image Segmentation?
10

 Fuzzy C-Means (FCM)
• Partition a finite collection of pixels into a
collection of "C" fuzzy clusters [22]
 Region Growing (RG)
• Group of pixels with similar properties to form a
region
• For similarity measure, difference between a
pixel's intensity value and the region's mean [23]
Image Segmentation Algorithms
11
22. M. Singha and K. Hemachandran, "Color Image Segmentation for Satellite Images", IJCSE, vol. 3(12), 2011.
23. M. Edman, "Segmentation Using a Region Growing Algorithm," Rensselaer Polytechnic Institute, 2007.

 Hill Climbing with K-Means (HKM)
• detects local maxima of clusters in global three-
dimensional color histogram of an image [28]
 Watershed (WS)
• It comes from geography
• It is that of a topographic relief which is flooded
by water
• Watershed lines being the divide lines of the
domains of attraction of rain falling over the
region [6]
Image Segmentation Algorithms
12
28. R. Vijayanandh and G. Balakrishnan, "Hill climbing Segmentation with Fuzzy C-Means Based Human Skin Region
Detection using Bayes Rule," EJSR, Vol. 76(1), pp. 95-107, 2012. 6. X. Han, Y. Fu and H. Zhang, "A Fast Two-Step
Marker-Controlled Watershed Image Segmentation Method," Proceedings of ICMA, pp. 1375-1380, 2012

Proposed Watershed Algorithm
• It can quickly calculate the
every region of the watershed
segmentation
• Image normalization
operation by Eq. 1
13

 Adaptive threshold determined by Eq. 2 and Eq. 3
based on Gray-threshold function
 N-dimensional convolution for smoothing image
 Adaptive masking operations by Eq. 4 and Eq. 5
14

 Impose Minima to create morphological process
image using Nucleus-masking (M2) on three color
channels
 Apply Watershed algorithm (Wn) on three color
channels
 Pixel labeling calculated by Ln = BWLABEL (Wn)
15

 Convert three channels into a RGB image for
visualizing labeled regions by Pn = label2rgb (Ln)
 R, G and B color channels (Pn) are added to generate
segmented image
16

 Applied canny edge detection method to detect
enclosed region boundary and remove all small
object from the combined three color channels
 The enclosed region boundary is superimposed on
original image in the final segmentation
17

 Peak Signal to Noise Ratio (PSNR) is
calculated between two images by Eq. 6 [40].
 Mean Square Error (MSE) is calculated pixel-
by-pixel by adding up the squared difference of all
the pixels and dividing by the total pixel count
using the Eq. 7 [40].
 Image Quality Measure (CQM) is based on
color transformation from RGB to YUV.
Quality Evaluation Metrics
1840. C. Mythili and V. Kavitha, "Color Image Segmentation using ERKFCM," IJCA, Vol. 41(20), pp. 21-28, 2012

 Reversible YUV Color Transformation (RCT) that is
created from the JPEG2000 standard in Eq. 8
 PSNR of each YUV color channel (Y, U and V) is
calculated separately
 CQM value is calculated using the Eq. 9 [43].
 Riesz-transform based Feature Similarity
Metric (RFSIM) is based on the human vision
system (HVS) perceives an image mainly according
to its low-level features
19
43. Y. YALMAN and Đ. ERTÜRK, "A new color image quality measure based on yuv transformation and psnr for
human vision system,“ Turkish Journal of Electrical Engineering & Computer Sciences, 2013, in press.

 Compute the similarity between two images f and g
 M1 and M2 is the result of edge detection performed
on f and g
 Then, the feature mask is defined as Eq. 10.
 Similarity between two feature maps fi (i = 1~5) and
gi at the corresponding location (x, y) is defined as
the Hilbert transform of a 1-D function in Eq. 11.
20

 To define similarity between feature maps fi and gi
by considering only key locations marked by mask M
and Hilbert transform of a 2-D function by Eq. 12
 RFSIM index computes between f and g image as
Eq. 13 [42]
 RFSIM range between [0, 1), the higher RFSIM
value indicates better image quality
21
42. L. Zhang, L. Zhang and X. Mou, "RFSIM: A Feature based image quality assessment metric using Riesz-
Transforms," Image Processing (ICIP), 17th IEEE International Conference, pp. 321-324, 2010.

 Visual Verification
• Comparative performance of the proposed MWS
method with four modified watershed methods
• Compared the results of the proposed algorithm
with three image segmentation algorithms
 Quantitative Verification
• Color image segmentation results with 20
different classes of images
• Performance of proposed method with three
different algorithms with respect to 5 IQA metrics
Results Analysis
22

23
33. C. Zhang, S. Zhang, J. Wu, S. Han, "An improved watershed algorithm for color image segmentation,“ I CCSEE, pp.
69-72, 2012. 35. S. Li, J. Xu, J. Ren and T. Xu, "A Color Image Segmentation Algorithm by Integrating Watershed with
Region Merging," RSKT, LNAI 7414, pp. 167–173, 2012.

24
7. H. Tan, Z. Hou, X. Li, R. Liu and W. Guo, "Improved watershed algorithm for color image segmentation," Proc. of
SPIE Vol. 7495 74952Z-(1-8). 9. L. Gao, S. Yang, J. Xia, S. Wang, J. Liang and Y. Qin, "New Marker-Based Watershed
Algorithm," TENCON 2006.

 A novel image segmentation method based on
adaptive threshold and masking operation with
watershed algorithm
 Compared the proposed MWS algorithm with four
modified watershed algorithms
 The results achieved using my technique ensure
accuracy and quality of the image in 20 different
classes of images in four segmentation algorithms
 Proposed method is less computational complexity,
which makes it appropriate for real-time application
 In future I am going to develop a robust algorithm
for the segmentation of color and video images
Conclusions
29

1) "Segmentation of Color Image using Adaptive
Thresholding and Masking with Watershed Algorithm,"
Presented at 2nd International Conference on
Informatics, Electronics & Vision (ICIEV), Dhaka
University, Bangladesh, ISBN: 978-1-4799-0399-3,
May 2013 (To appear in IEEE Xplore).
2) "A version of watershed algorithm for color image
segmentation," AIUB Journal of Science and
Engineering (AJSE), Bangladesh, Vol. 12(1), 2013
(accepted).
List of Publication related to this thesis
30

[6] X. Han, Y. Fu and H. Zhang, "A Fast Two-Step Marker-Controlled Watershed Image Segmentation Method,"
Proceedings of ICMA, pp. 1375-1380, 2012.
[7] H. Tan, Z. Hou, X. Li, R. Liu and W. Guo, "Improved watershed algorithm for color image segmentation,"
Proc. of SPIE Vol. 7495 74952Z-(1-8).
[9] L. Gao, S. Yang, J. Xia, S. Wang, J. Liang, and Y. Qin, "New Marker-Based Watershed Algorithm," TENCON
2006.
[11] R. Gonzalez and R. Woods, “Digital Image Processing,” 3rd edition, Pearson Prentice Hall, 2007.
[13] H. Vankayalapati, "Evaluation of wavelet based linear subspace techniques for face recognition," Alpen-
Adria University and Institute for Smart System-Technologies, Klagenfurt, 2008.
[22] M. Singha and K. Hemachandran, "Color Image Segmentation for Satellite Images", IJCSE, vol. 3(12), 2011.
[23] M. Edman, "Segmentation Using a Region Growing Algorithm," Rensselaer Polytechnic Institute, 2007.
[28] R. Vijayanandh and G. Balakrishnan, "Hill climbing Segmentation with Fuzzy C-Means Based Human Skin
Region Detection using Bayes Rule," EJSR, Vol. 76(1), pp. 95-107, 2012.
[33] C. Zhang, S. Zhang, J. Wu, S. Han, "An improved watershed algorithm for color image segmentation,"
International Conference on Computer Science and Electronics Engineering (ICCSEE), pp. 69-72, 2012.
[35] S. Li, J. Xu, J. Ren, and T. Xu, "A Color Image Segmentation Algorithm by Integrating Watershed with
Region Merging," RSKT, LNAI 7414, pp. 167–173, 2012.
[40] C. Mythili and V. Kavitha, "Color Image Segmentation using ERKFCM," International Journal of Computer
Applications (IJCA), Vol. 41(20), pp. 21-28, 2012.
[42] L. Zhang, L. Zhang, and X. Mou, "RFSIM: A Feature based image quality assessment metric using Riesz-
Transforms," Image Processing (ICIP), 17th IEEE International Conference, pp. 321-324, 2010.
[43] Y. YALMAN and Đ. ERTÜRK, "A new color image quality measure based on yuv transformation and psnr
for human vision system," Turkish Journal of Electrical Engineering & Computer Sciences, 2013, in press.
Some Important References
31

A version of watershed algorithm for color image segmentation

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to A version of watershed algorithm for color image segmentation (20)

More from Habibur Rahman (10)

Recently uploaded (20)

A version of watershed algorithm for color image segmentation