Skip to main content

Table 2 The PSNR(dB)/SSIM values of reconstructed SR images in 3 datasets

From: Single image super resolution based on multi-scale structure and non-local smoothing

Upscale ratio

Dataset

Bicubic (PSNR/SSIM)

SC [11] (PSNR/SSIM)

A+ [13] (PSNR/SSIM)

SRCNNb [19] (PSNR/SSIM)

SRCNNg [19] (PSNR/SSIM)

NCSR [17] (PSNR/SSIM)

Proposed (PSNR/SSIM)

x3

Set5

26.48/0.79

26.72/0.80

26.72/0.81

26.73/0.81

29.67/0.85

33.19/0.91

33.40/0.92

 

Set9

25.20/0.70

25.49/0.72

25.51/0.73

25.51/0.73

26.92/0.76

29.86/0.85

30.06/0.85

 

Set14

24.72/0.68

24.92/0.70

24.91/0.70

24.93/0.70

27.20/0.76

29.38/0.82

29.51/0.83

x4

Set5

24.75/0.73

24.67/0.74

24.52/0.74

24.43/0.74

28.54/0.81

30.65/0.87

31.27/0.88

 

Set9

23.91/0.65

23.90/0.66

23.78/0.66

23.72/0.66

26.13/0.72

27.71/0.79

28.24/0.80

 

Set14

23.39/0.62

23.31/0.63

23.18/0.63

23.06/0.63

26.22/0.72

27.50/0.76

27.91/0.76

  1. The PSNR/SSIM of the reconstructed images from our proposed method (the numbers in boldface) outperforms other SOTA conventional SR methods and the DL based SR methods