Table 3 Quantitative comparison of proposed algorithm with existing state-of-the-art machine learning-based image defogging
From: Segmentation-based image defogging using modified dark channel prior
Quality Metrics | Dehaze-Net [37] | MSCNN [36] | AoD-Net [38] | Proposed I | Proposed II |
|---|---|---|---|---|---|
Image 1 | |||||
SSIM | 0.7362 | 0.8040 | 0.8245 | 0.86233 | 0.8916 |
PSNR | 19.0583 | 17.401 | 20.9252 | 19.4014 | 20.9254 |
NIQE | 2.6246 | 2.454 | 2.2992 | 3.1089 | 3.0930 |
BRISQUE | 20.2741 | 27.4012 | 18.9295 | 18.4556 | 18.5600 |
Image 2 | |||||
SSIM | 0.863 | 0.8645 | 0.9432 | 0.93417 | 0.9032 |
PSNR | 19.5769 | 19.7853 | 21.7582 | 22.477 | 20.5093 |
NIQE | 3.9958 | 4.0555 | 20.6351 | 4.0373 | 4.3324 |
BRISQUE | 22.391 | 24.9614 | 18.5274 | 17.7474 | 31.0225 |
Image 3 | |||||
SSIM | 0.8859 | 0.8635 | 0.8623 | 0.91604 | 0.9184 |
PSNR | 24.4493 | 22.1310 | 25.5093 | 22.9655 | 25.0210 |
NIQE | 3.5345 | 3.2623 | 4.1446 | 3.2526 | 4.3324 |
BRISQUE | 17.4855 | 28.6885 | 18.2809 | 16.8705 | 22.5321 |