Normal Inverse Gaussian Model-Based Image Denoising in the NSCT Domain
Table 2
Denoising results (PSNR in dB) for noisy versions of seven images, obtained with BM3D, Ram’s method, and the proposed scheme. The best result for each image and noise level is displayed in boldface.
Image
Method
σ/PSNR
5/34.16
10/28.14
15/24.61
20/22.11
25/20.18
50/14.16
75/10.63
100/8.14
Boats
Ram’s
37.07
33.68
31.91
30.69
29.72
26.72
25.03
23.92
BM3D
37.28
33.92
32.14
30.88
29.91
26.64
24.96
23.74
Proposed
37.50
34.07
32.25
31.00
30.05
27.23
25.67
24.57
Fingerprint
Ram’s
36.17
32.00
29.83
28.37
27.33
24.10
22.44
21.46
BM3D
36.51
32.46
30.28
28.81
27.70
24.36
22.68
21.33
Proposed
36.36
32.62
30.58
29.16
28.10
24.91
23.12
21.87
Man
Ram’s
37.38
33.67
31.77
30.48
29.55
26.74
25.17
24.15
BM3D
37.82
33.98
31.93
30.59
29.62
26.59
25.10
23.97
Proposed
37.78
34.17
32.29
31.04
30.11
27.42
25.90
24.82
Couple
Ram’s
37.12
33.60
31.74
30.46
29.42
26.39
24.69
23.50
BM3D
37.52
34.04
32.11
30.76
29.72
26.38
24.63
23.37
Proposed
37.36
33.80
31.90
30.62
29.66
26.83
25.27
24.19
Hill
Ram’s
36.92
33.48
31.77
30.64
29.76
27.17
25.61
24.58
BM3D
37.14
33.62
31.86
30.72
29.85
27.08
25.58
24.45
Proposed
37.31
33.93
32.21
31.07
30.22
27.72
26.31
25.29
House
Ram’s
38.76
35.82
34.38
33.33
32.55
29.77
27.68
26.20
BM3D
39.83
36.71
34.94
33.77
32.86
29.37
27.20
25.50
Proposed
39.44
36.13
34.27
32.98
32.00
28.88
27.00
25.64
Peppers
Ram’s
37.62
34.22
32.33
31.02
29.92
26.72
24.75
23.32
BM3D
38.12
34.68
32.70
31.29
30.16
26.41
24.48
22.91
Proposed
37.64
34.12
32.18
30.83
29.80
26.65
24.81
23.51
Bold number indicates best performing method in separate column.