Research Article
Convergence Analysis of a Modified Forward-Backward Splitting Algorithm for Minimization and Application to Image Recovery
Table 2
The results of deblurred images for each algorithm.
| Blur types | Methods | Fig(a) | Fig(b) | PSNR | SSIM | PSNR | SSIM |
| Out-of-focus blur (disk) with radius (see Figure 2) | FISTA-BT | 34.1671 | 0.8951 | 44.6086 | 0.9855 | FISTA-CN | 36.3911 | 0.9327 | 45.8992 | 0.9888 | NAGA | 35.6176 | 0.9222 | 45.4853 | 0.9878 | NMFBS | 40.1968 | 0.9728 | 47.3828 | 0.9916 | Gaussian blur of the filter size with standard deviation (see Figure 3) | FISTA-BT | 34.8432 | 0.9631 | 45.8546 | 0.9906 | FISTA-CN | 36.5138 | 0.9735 | 47.0315 | 0.9926 | NAGA | 35.9527 | 0.9703 | 46.6534 | 0.9920 | NMFBS | 38.9891 | 0.9836 | 48.4820 | 0.9946 | Motion blur specified with the motion length of pixels and motion orientation (see Figure 4) | FISTA-BT | 40.5480 | 0.9548 | 48.6089 | 0.9940 | FISTA-CN | 44.0301 | 0.9782 | 50.0219 | 0.9955 | NAGA | 42.7566 | 0.9709 | 49.6478 | 0.9952 | NMFBS | 50.5468 | 0.9949 | 51.3704 | 0.9967 |
|
|