Laser & Optoelectronics Progress, Volume. 59, Issue 4, 0415007(2022)
Image Defogging Algorithm Based on Generative Adversarial Network
Figures & Tables(12)
Fig. 6. Errors in 20-layer and 56-layer network training. (a) Training error; (b) test error
Fig. 8. Experimental results of different algorithms on foggy trees. (a) Original picture; (b) AODNet algorithm; (c) DehazeNet algorithm; (d) FFANet algorithm; (e) GCANet algorithm; (f) improved CycleGAN algorithm
Fig. 9. Experimental results of different algorithms on foggy scenes. (a) Original picture; (b) AODNet algorithm; (c) DehazeNet algorithm; (d) FFANet algorithm; (e) GCANet algorithm; (f) improved CycleGAN algorithm
Table 1. Size information of generator
View tableTable 1. Size information of generator
Layer name Kernel size Stride Conv layer 1 7×7 1 Conv layer 2 3×3 2 Conv layer 3 3×3 2 Conv layer 4 7×7 1 Upsample layer 1 3×3 2 Upsample layer 2 3×3 2
Table 2. Evaluation indexes of Fig.8
View tableTable 2. Evaluation indexes of Fig.8
Algorithm PSNR /dB SSIM Information entropy /bit AODNet 28.6010 0.7757 6.8782 DehazeNet 28.7951 0.5701 7.3590 FFANet 28.8670 0.7751 7.1810 GCANet 28.8296 0.6441 7.5685 CycleGAN 28.8946 0.8090 7.6027
Table 3. Evaluation indexes of Fig.9
View tableTable 3. Evaluation indexes of Fig.9
Algorithm PSNR /dB SSIM Information entropy /bit AODNet 29.1986 0.7694 7.1634 DehazeNet 29.2523 0.7075 7.4437 FFANet 29.3011 0.7880 7.2855 GCANet 28.9877 0.7360 7.2636 CycleGAN 29.4215 0.8271 7.4751
Tools
Get Citation
Copy Citation Text
Weifeng Zhong, Jing Zhao. Image Defogging Algorithm Based on Generative Adversarial Network[J]. Laser & Optoelectronics Progress, 2022, 59(4): 0415007
Paper Information
Category: Machine Vision
Received: Jul. 14, 2021
Accepted: Sep. 13, 2021
Published Online: Feb. 15, 2022
The Author Email: Weifeng Zhong (zhongweifeng@hrbust.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20