Dataset | Scale | Bicubic | SF | Zeyde et al. | GR | ANR | NE+LS | NE+NNLS | NE+LLE | Our A+ (0.5mil) | Our A+ | Our A+ (only 16 atoms!) | |
Set5 | x2 | 33.66 | 35.63 | 35.78 | 35.13 | 35.83 | 35.66 | 35.43 | 35.77 | 36.37 | 36.55 (+0.72 over ANR) | 35.95 | |
Set5 | x3 | 30.39 | 31.27 | 31.90 | 31.41 | 31.92 | 31.78 | 31.60 | 31.84 | 32.39 | 32.59 (+0.67 over ANR) | 32.07 | |
Set5 | x4 | 28.42 | 28.94 | 29.69 | 29.34 | 29.69 | 29.55 | 29.47 | 29.61 | 30.04 | 30.29 (+0.60 over ANR) | 29.83 | |
Set14 | x2 | 30.23 | 31.04 | 31.81 | 31.36 | 31.80 | 31.69 | 31.55 | 31.76 | 32.13 | 32.28 (+0.48 over ANR) | 31.91 | |
Set14 | x3 | 27.54 | 28.20 | 28.67 | 28.31 | 28.65 | 28.59 | 28.44 | 28.60 | 28.97 | 29.13 (+0.48 over ANR) | 28.71 | |
Set14 | x4 | 26.00 | 26.25 | 26.88 | 26.60 | 26.85 | 26.81 | 26.72 | 26.81 | 27.10 | 27.32 (+0.47 over ANR) | 26.99 | |
B100 | x2 | 29.32 | 30.35 | 30.40 | 30.23 | 30.44 | 30.36 | 30.27 | 30.41 | 30.68 | 30.78 (+0.34 over ANR) | 30.53 | |
B100 | x3 | 27.15 | 27.76 | 27.87 | 27.70 | 27.89 | 27.83 | 27.73 | 27.85 | 28.08 | 28.18 (+0.29 over ANR) | 27.94 | |
B100 | x4 | 25.92 | 26.19 | 26.51 | 26.37 | 26.51 | 26.45 | 26.41 | 26.47 | 26.65 | 26.77 (+0.26 over ANR) | 26.58 | |
[43] | R. Timofte, R. Rothe, and L. Van Gool. Seven ways to improve example-based single image super resolution. In IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR 2016), June 2016, US. [ bib | pdf | Supplementary material | L20 dataset ] |