Research Article
Pixelwise Estimation of Signal-Dependent Image Noise Using Deep Residual Learning
Table 2
Standard deviation of error on simulated homogeneous noise (dB).
| Dataset | Noise level | Pyatykh [24] | Liu [25] | Chen [7] | DRNE |
| Kodak (24 images) | | 0.45 | 0.30 | 0.22 | 0.09 | | 0.51 | 0.26 | 0.15 | 0.15 | | 1.00 | 0.49 | 0.36 | 0.30 | | 1.46 | 0.67 | 0.55 | 0.50 | | 1.91 | 0.86 | 0.75 | 0.71 | | 2.31 | 1.07 | 0.95 | 0.96 |
| McMaster (18 images) | | 0.16 | 0.10 | 0.71 | 0.24 | | 0.17 | 0.18 | 0.28 | 0.11 | | 0.52 | 0.46 | 0.34 | 0.32 | | 0.87 | 0.76 | 0.67 | 0.62 | | 1.11 | 1.09 | 1.03 | 0.94 | | 1.35 | 1.42 | 1.37 | 1.28 |
| BSD500 (500 images) | | 0.77 | 0.26 | 0.21 | 0.07 | | 1.12 | 1.31 | 0.08 | 0.11 | | 0.80 | 0.42 | 0.22 | 0.26 | | 1.25 | 0.61 | 0.41 | 0.47 | | 1.71 | 0.85 | 0.64 | 0.72 | | 2.17 | 1.11 | 0.91 | 1.00 |
|
|
Bold fonts denote the best performance and italics denotes the second best performance.
|