Image Denoising Using Nonlocal Means with Shape-Adaptive Patches and New Weights
Table 2
The SSIM results by different denoising methods. In each cell, the results of the four denoising methods are presented in the following order: top left, NLM [21]; top right, NLM-SAP [29]; bottom left, BM3D-SAPCA [41]; bottom right, our proposed denoising method.
σ
20
35
50
100
C. man
0.840
0.845
0.763
0.773
0.714
0.706
0.541
0.493
0.886
0.854
0.827
0.801
0.787
0.758
0.643
0.587
House
0.831
0.849
0.775
0.787
0.733
0.715
0.553
0.531
0.876
0.870
0.838
0.836
0.807
0.788
0.676
0.639
Peppers
0.840
0.863
0.759
0.794
0.703
0.730
0.518
0.560
0.886
0.877
0.833
0.826
0.792
0.772
0.666
0.631
Lena
0.830
0.918
0.764
0.858
0.724
0.801
0.574
0.644
0.880
0.879
0.837
0.819
0.801
0.772
0.674
0.652
Barbara
0.861
0.925
0.779
0.843
0.711
0.768
0.510
0.608
0.912
0.894
0.863
0.815
0.811
0.737
0.600
0.531
Boats
0.786
0.876
0.689
0.789
0.625
0.720
0.471
0.561
0.828
0.810
0.761
0.737
0.708
0.679
0.578
0.531
Man
0.793
0.871
0.692
0.785
0.625
0.721
0.473
0.582
0.840
0.825
0.763
0.745
0.710
0.684
0.579
0.555
Hill
0.765
0.851
0.659
0.757
0.589
0.694
0.448
0.572
0.809
0.790
0.729
0.709
0.675
0.648
0.549
0.535
Average
0.818
0.874
0.735
0.798
0.678
0.731
0.511
0.568
0.864
0.849
0.806
0.786
0.761
0.729
0.620
0.582
The bold values represent the best results among the four methods, which has been explained in the second paragraph of Section 4.
