Segmentation Results through FCM clustering with MATLAB program
-
Upload
aditya-pavan-kumar -
Category
Engineering
-
view
32 -
download
3
Transcript of Segmentation Results through FCM clustering with MATLAB program
![Page 1: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/1.jpg)
1
SEGMENTATION RESULTS THROUGH FCM CLUSTERING
![Page 2: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/2.jpg)
2
Notes
• The thresholding factor of FCM is varied and segmentation outputs are observed for different values of “Thresholding factor”.
• Jaccard index is calculated for each segmentation output with respect to the provided “Ground truth image”.
![Page 3: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/3.jpg)
3
Sample 1
GROUND TRUTH IMAGE
![Page 4: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/4.jpg)
4
FCM SEGMENTATION RESULTS
THRESHOLDING FACTOR U1(V)>0.4
JACCARD INDEX = 0.1362
THRESHOLDING FACTOR U1(V)>0.6
JACCARD INDEX=0.2859
![Page 5: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/5.jpg)
5
THRESHOLDING FACTOR U1(V)>0.8
THRESHOLDING FACTOR U1(V)>0.99
JACCARD INDEX=0.8350
JACCARD INDEX=0.2960
![Page 6: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/6.jpg)
6
SAMPLE 2
GROUND TRUTH IMAGE
![Page 7: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/7.jpg)
7
FCM SEGMENTATION RESULTS
THRESHOLDING FACTOR U1(V)>0.4
THRESHOLDING FACTOR U1(V)>0.6
JACCARD INDEX= 0.2498
JACCARD INDEX=0.2898
![Page 8: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/8.jpg)
8
THRESHOLDING FACTOR U1(V)>0.8
THRESHOLDING FACTOR U1(V)>0.99
JACCARD INDEX=0.5031
JACCARD INDEX=0.6309
![Page 9: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/9.jpg)
9
SAMPLE 3
GROUND TRUTH IMAGE
![Page 10: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/10.jpg)
10
FCM SEGMENTED RESULTS
THRESHOLDING FACTOR U1(V)>0.4
THRESHOLDING FACTOR U1(V)>0.6
JACCARD INDEX=0.3750
JACCARD INDEX=0.4443
![Page 11: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/11.jpg)
11
THRESHOLDING FACTOR U1(V)>0.8
THRESHOLDING FACTOR U1(V)>0.99
JACCARD INDEX=0.0929
JACCARD INDEX=0
![Page 12: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/12.jpg)
12
SAMPLE 4
GROUND TRUTH IMAGE
![Page 13: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/13.jpg)
13
FCM SEGMENTED RESULTS
THRESHOLDING FACTOR U1(V)>0.4
THRESHOLDING FACTOR U1(V)>0.6
JACCARD INDEX=0.3817
JACCARD INDEX=0.7398
![Page 14: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/14.jpg)
14
THRESHOLDING FACTOR U1(V)>0.8
THRESHOLDING FACTOR U1(V)>0.99
JACCARD INDEX=0.8120
JACCARD INDEX=0.1684
![Page 15: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/15.jpg)
15
SAMPLE 5
GROUND TRUTH IMAGE
![Page 16: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/16.jpg)
16
FCM SEGMENTED RESULTS
THRESHOLDING FACTOR U1(V)>0.4
THRESHOLDING FACTOR U1(V)>0.6
JACCARD INDEX=0.3801
JACCARD INDEX=0.7525
![Page 17: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/17.jpg)
17
THRESHOLDING FACTOR U1(V)>0.8
THRESHOLDING FACTOR U1(V)>0.99
JACCARD INDEX=0.6881
JACCARD INDEX=0.0298
![Page 18: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/18.jpg)
18
Program for Segmentation and Jaccard Index Calculation
![Page 19: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/19.jpg)
19
Program(continued)
![Page 20: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/20.jpg)
20
Observations
• The Ideal threshold value for best result is different for different sample images.
• The Image has few unwanted features like shadows, hair, other little impressions etc., which have appeared in our “Segmentation Outputs ”.
![Page 21: Segmentation Results through FCM clustering with MATLAB program](https://reader035.fdocuments.net/reader035/viewer/2022070600/58d108c01a28ab823e8b54f7/html5/thumbnails/21.jpg)
21
Observations(continued)
• Preprocessing the image before segmenting it using FCM clustering algorithm is required to improve the Jaccard Index.
• The Jaccard Index is a better similarity measure compared to spatial overlap index as it compares only the white regions of the images i.e., see sample 3, result 4.