Table 1 Quantitative comparisons with state-of-the-art methods on three datasets, including D_Denoise, D_Restore, and D_Synthetic, measured by average PSNR and SSIM metrics

Raw Image

8.23

0.3689

8.01

0.3566

15.14

0.1718

DeepRes⁵⁶

13.77

0.5969

12.90

0.5765

21.59

0.8397

DnCNN⁴⁷

13.63

0.6245

13.31

0.6044

18.81

0.7591

IDCCW³⁸

15.06

0.6580

14.49

0.5998

21.62

0.8553

SinGAN³⁷

17.98

0.8129

18.05

0.8238

22.64

0.8974

Noise2same⁵⁸

13.12

0.6103

12.57

0.5931

18.86

0.7624

InvDN⁵⁷

17.22

0.7980

16.97

0.7630

22.18

0.8840

Wavelet filtering⁶⁰

9.79

0.3850

9.55

0.3907

16.22

0.2459

Bilateral filtering⁶⁰

10.42

0.4031

10.30

0.4129

16.62

0.2557

BM3D⁵⁹

10.58

0.4259

10.22

0.4142

20.06

0.8368

DiffACR³⁹

16.84

0.8248

16.65

0.8254

21.76

0.8864

Our Method

19.32

0.8663

18.83

0.8732

22.69

0.8942