Research Article
Speckle Noise Removal Model Based on Diffusion Equation and Convolutional Neural Network
Table 2
PSNR results of DnCNN on different training sets.
| Test image | I | II | III | IV | V | VI | VII | VIII | Average |
| Noise level σ = 15 | DnCNN (single) DnCNN (mix) | 32.2732.01 | 34.64 | 32.62 | 33.37 | 31.72 | 33.29 | 35.02 | 33.39 | 33.29 | 34.22 | 32.11 | 33.12 | 31.54 | 32.93 | 34.83 | 33.01 | 32.97 | Noise level σ = 25 | DnCNN (single) DnCNN (mix) | 29.51 | 32.48 | 30.28 | 30.92 | 29.43 | 30.48 | 33.36 | 30.89 | 30.91 | 29.30 | 32.12 | 29.34 | 30.12 | 29.08 | 30.21 | 33.01 | 30.34 | 30.44 |
|
|