Chinese Journal of Lasers, Volume. 47, Issue 8, 814003(2020)
Terahertz Holographic Reconstructed Image Segmentation Based on Optimized Region Growth by Evolutionary Algorithm
Figures & Tables(31)
Fig. 3. Gear images. (a) Original image; (b) standard image; (c) preprocessed image; (d) seed image
Fig. 4. Segmentation results of gear images obtained by different algorithms for f1. (a) RG-GA; (b) RG-DE
Fig. 5. Difference images of segmentation results obtained by different algorithms for f1. (a) RG-GA; (b) RG-DE
Fig. 6. Segmentation results of gear images obtained by different algorithms for f2. (a) RG-GA; (b) RG-DE
Fig. 7. Difference images of segmentation results obtained by different algorithms for f2. (a) RG-GA; (b) RG-DE
Fig. 8. Shim images. (a) Original image; (b) standard image; (c) preprocessed image; (d) seed image
Fig. 9. Segmentation results of shim images obtained by different algorithms for f1. (a) RG-GA; (b) RG-DE
Fig. 10. Difference images of segmentation results obtained by different algorithms for f1. (a) RG-GA; (b) RG-DE
Fig. 11. Segmentation results of shim images obtained by different algorithms for f2. (a) RG-GA; (b) RG-DE
Fig. 12. Difference images of segmentation results obtained by different algorithms for f2. (a) RG-GA; (b) RG-DE
Fig. 14. Segmentation results of Z images obtained by different algorithms for f1. (a) RG-GA; (b) RG-DE
Fig. 15. Difference images of segmentation results obtained by different algorithms for f1. (a) RG-GA; (b) RG-DE
Fig. 16. Segmentation results of Z images by obtained different algorithms for f2. (a) RG-GA; (b) RG-DE
Fig. 17. Difference images of segmentation results obtained by different algorithms for f2. (a) RG-GA; (b) RG-DE
Fig. 18. Simulation images. (a) Simulation image without noise; (b) simulation image with noise; (c) standard image
Fig. 19. Segmentation results of simulation images obtained by different algorithms for f1. (a) RG-GA; (b) RG-DE
Fig. 20. Difference images of segmentation results obtained by different algorithms for f1. (a) RG-GA; (b) RG-DE
Fig. 21. Segmentation results of simulation images obtained by different algorithms for f2. (a) RG-GA; (b) RG-DE
Fig. 22. Difference images of segmentation results obtained by different algorithms for f2. (a) RG-GA; (b) RG-DE
Table 1. Parameter setting of GA and DE
View tableTable 1. Parameter setting of GA and DE
Algorithm Number of variables Population size Maximum number of iterations Crossover probability Mutation probability Mutation step GA 1 5 50 0.9 0.2 none DE 1 5 50 0.9 none 0.5
Table 2. Performance comparison between RG-GA and RG-DE in gear images for f1
View tableTable 2. Performance comparison between RG-GA and RG-DE in gear images for f1
Algorithm MSSIM Time /s Number of iterations T Number of over segmentation pixels Number of under segmentation pixels Optimum fitness value RG-GA 0.92 7.05 20 1.53 10 406 -0.052 RG-DE 0.92 1.62 15 1.53 10 406 -0.052
Table 3. Performance comparison between RG-GA and RG-DE in gear images for f2
View tableTable 3. Performance comparison between RG-GA and RG-DE in gear images for f2
Algorithm MSSIM Time /s Number of iterations T Number of over segmentation pixels Number of under segmentation pixels Optimum fitness value RG-GA 0.938 4.55 13 7.90 20 290 1.07 RG-DE 0.942 2.10 14 13.77 25 261 1.06
Table 4. Performance comparison between RG-GA and RG-DE in shim images for f1
View tableTable 4. Performance comparison between RG-GA and RG-DE in shim images for f1
Algorithm MSSIM Time /s Number of iterations T Number of over segmentation pixels Number of under segmentation pixels Optimum fitness value RG-GA 0.68 2.87 5 40.80 1401 62 -0.007 RG-DE 0.67 0.12 5 51.00 1440 61 -0.009
Table 5. Performance comparison between RG-GA and RG-DE in shim images for f2
View tableTable 5. Performance comparison between RG-GA and RG-DE in shim images for f2
Algorithm MSSIM Time /s Number of iterations T Number of over segmentation pixels Number of under segmentation pixels Optimum fitness value RG-GA 0.91 3.75 3 19.38 36 118 2.56 RG-DE 0.91 1.70 3 19.38 36 118 2.56
Table 6. Performance comparison between RG-GA and RG-DE in Z images for f1
View tableTable 6. Performance comparison between RG-GA and RG-DE in Z images for f1
Algorithm MSSIM Time /s Number of iterations T Number of over segmentation pixels Number of under segmentation pixels Optimum fitness value RG-GA 0.99 2.40 3 66.3 7 11 -0.063 RG-DE 0.99 1.21 2 35.7 4 11 -0.064
Table 7. Performance comparison between RG-GA and RG-DE in Z images for f2
View tableTable 7. Performance comparison between RG-GA and RG-DE in Z images for f2
Algorithm MSSIM Time /s Number of iterations T Number of over segmentation pixels Number of under segmentation pixels Optimum fitness value RG-GA 0.99 3.70 3 63.75 7 11 1.524 RG-DE 0.99 1.70 2 38.25 4 12 1.521
Table 8. Performance comparison between RG-GA and RG-DE in simulation images for f1
View tableTable 8. Performance comparison between RG-GA and RG-DE in simulation images for f1
Algorithm MSSIM Time /s Number of iterations T Number of over segmentation pixels Number of under segmentation pixels Optimum fitness value RG-GA 0.92 9.83 22 3.06 12 345 -0.0188 RG-DE 0.92 3.30 16 2.60 12 345 -0.0188
Table 9. Performance comparison between RG-GA and RG-DE in simulation images for f2
View tableTable 9. Performance comparison between RG-GA and RG-DE in simulation images for f2
Algorithm MSSIM Time /s Number of iterations T Number of over segmentation pixels Number of under segmentation pixels Optimum fitness value RG-GA 0.93 6.50 16 3.57 13 328 1.08 RG-DE 0.93 3.57 14 3.57 13 328 1.08
Tools
Get Citation
Copy Citation Text
Wang Yutong, Li Qi. Terahertz Holographic Reconstructed Image Segmentation Based on Optimized Region Growth by Evolutionary Algorithm[J]. Chinese Journal of Lasers, 2020, 47(8): 814003
Paper Information
Category: terahertz technology
Received: Feb. 4, 2020
Accepted: --
Published Online: Aug. 17, 2020
The Author Email: Yutong Wang (hit_wyt@sina.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20