Table 2 The quantitative results of Gaussian denoising with different noise level on AWLD dataset (PSNR/SSIM). Bolditalic number represents the best result, while bold number indicates the second-best result.
From: Wild horseshoe crab image denoising based on CNN-transformer architecture
Method | sigma = 15 | sigma = 25 | sigma = 50 | AVG | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
UI | UTI | TI | UI | UTI | TI | UI | UTI | TI | ||
AP-BSN | 37.19/0.9502 | 36.15/0.9496 | 35.82/0.9579 | 37.09/0.9487 | 35.27/0.9431 | 34.53/0.9368 | 35.72/0.9452 | 34.51/0.9361 | 31.92/0.8991 | 35.36/0.9407 |
DDT | 38.68/0.9671 | 38.56/0.9617 | 36.69/0.9501 | 37.29/0.9503 | 37.39/0.9493 | 36.44/0.9508 | 36.29/0.9489 | 36.74/0.9516 | 36.03/0.9453 | 37.12/0.9528 |
MIRNetv2 | 37.19/0.9496 | 35.19/0.9437 | 35.28/0.9441 | 36.54/0.9504 | 34.21/0.9364 | 34.29/0.9359 | 35.46/0.9455 | 32.98/0.9028 | 32.88/0.9011 | 34.89/0.9344 |
MM-BSN | 38.26/0.9613 | 36.19/0.9496 | 35.38/0.9446 | 37.41/0.9487 | 35.65/0.9443 | 34.18/0.9347 | 36.21/0.9432 | 32.93/0.9027 | 33.95/0.9104 | 35.57/0.9377 |
MPRNet | 38.19/0.9608 | 35.04/0.9428 | 36.17/0.9495 | 37.94/0.9525 | 34.26/0.9360 | 35.91/0.9461 | 36.64/0.9458 | 31.01/0.8922 | 34.25/0.9348 | 35.49/0.9401 |
SwinIR | 35.47/0.9561 | 33.57/0.9166 | 35.49/0.9537 | 34.18/0.9355 | 32.54/0.9001 | 34.37/0.9361 | 33.57/0.9126 | 30.04/0.8803 | 32.95/0.9008 | 33.58/0.9213 |
DDPG | 38.34/0.9618 | 32.94/0.9012 | 37.58/0.9501 | 37.15/0.9487 | 31.27/0.8946 | 36.04/0.9488 | 35.72/0.9464 | 30.12/0.8801 | 35.52/0.9451 | 34.96/0.9308 |
Restormer | 38.57/0.9635 | 37.06/0.9492 | 38.65/0.9633 | 36.52/0.9512 | 37.63/0.9508 | 37.07/0.9523 | 35.91/0.9468 | 36.16/0.9431 | 36.14/0.9460 | 37.08/0.9518 |
Ours | 39.26/0.9789 | 38.27/0.9603 | 39.12/0.9684 | 37.53/0.9526 | 37.25/0.9495 | 37.34/0.9531 | 36.48/0.9498 | 36.83/0.9521 | 37.02/0.9511 | 37.83/0.9573 |
