Table 1.
Average SSIM and PSNR of the simulated datasets, and the best results are shown in bold.
| Dataset | Metrics | Methods | |||
|---|---|---|---|---|---|
| Blurry | Blind Deconv | Wiener | CycleGAN | ||
| A | SSIM | 0.906 | 0.901 | 0.918 | 0.929 |
| PSNR (dB) | 26.97 | 27.52 | 27.92 | 29.77 | |
| B | SSIM | 0.861 | 0.796 | 0.845 | 0.886 |
| PSNR (dB) | 25.99 | 21.73 | 24.36 | 27.96 | |
| C | SSIM | 0.771 | 0.793 | 0.765 | 0.798 |
| PSNR (dB) | 22.07 | 21.58 | 22.24 | 24.27 | |
A represents the dataset that only adds Gaussian blur. B represents the dataset that adds Gaussian blur and the synthetic fog. C represents the dataset that adds hybrid blur (Gaussian blur, motion blur, out-of-focus blur) and the synthetic fog.