Table 4.
The average running time of a 500 × 500 image in McMaster and the number and size (for double-precision floating-point format) of parameters for CNN-based models. The bolded two methods in each section are those that achieve better performance among all algorithms.
| Algorithm | Running Time (s) | Parameters | |
|---|---|---|---|
| Number | Size (MB) | ||
| AHD [48] | 0.48 | - | - |
| DLMMSE [49] | 234.78 | - | - |
| RI [5] | 0.16 | - | - |
| MLRI [6] | 0.20 | - | - |
| ARI [7] | 3.66 | - | - |
| Tan [11] | 0.42 | 528,518 | 2.02 |
| Kokkinos [13] | 0.87 | 380,356 | 1.45 |
| Cui [14] | 1.19 | 1,793,032 | 6.84 |
| Ours (L1) | 0.14 | 11,786 | 0.04 |
| Ours (L2) | 0.17 | 46,537 | 0.18 |
| Ours (L3) | 0.24 | 183,628 | 0.70 |