The Scientific World Journal / 2014 / Article / Tab 5

Research Article

A “Salt and Pepper” Noise Reduction Scheme for Digital Images Based on Support Vector Machines Classification and Regression

Table 5

Table results in PSNR and MSSIM for SVM methods. The model used in all cases is with and . represents the mean and represents the standard deviation obtained using 10 different noise patterns.
(a)

PSNR results: Albert
Noise % SVM-M1 SVM-M2 SVM-R

10 37.806326 0.081000 37.634073 0.095187 37.806326 0.081000
20 34.012864 0.077311 34.247895 0.100104 34.012864 0.077311
30 31.334875 0.075222 32.111316 0.096603 31.334875 0.075222
40 28.782425 0.080101 30.524735 0.041627 28.782425 0.080101
50 27.047406 0.054255 29.098885 0.054799 27.047406 0.054255
60 24.689836 0.179538 27.809312 0.058029 24.689836 0.179538
70 22.478286 0.124136 26.584962 0.067960 22.478286 0.124136
80 15.891552 0.149735 25.358707 0.047150 15.891552 0.149735
90 8.109299 0.022287 23.885132 0.045141 8.109299 0.022287

(b)

PSNR results: Lenna
Noise % SVM-M1 SVM-M2 SVM-R

10 40.210190 0.220927 40.647102 0.137564 40.483306 0.123692
20 35.529780 0.412996 36.987532 0.152243 37.014919 0.148138
30 32.294158 0.230317 34.537564 0.092911 34.751509 0.085640
40 28.576325 0.381850 32.600946 0.062708 32.823432 0.055126
50 26.726578 0.183652 30.849589 0.072874 31.000031 0.114136
60 23.778253 0.182517 29.311999 0.071502 29.154773 0.102010
70 20.483630 0.121690 27.827742 0.055517 27.053785 0.074279
80 12.017827 0.171127 26.298444 0.065581 24.520623 0.047364
90 4.453599 0.014529 24.109759 0.068257 20.390460 0.101721

(c)

MSSIM results: Albert
Noise % SVM-M1 SVM-M2 SVM-R

10 0.970160 0.000260 0.970334 0.000257 0.969802 0.000227
20 0.932638 0.000469 0.936817 0.000381 0.936707 0.000548
30 0.884381 0.000755 0.899063 0.000610 0.900619 0.000888
40 0.821351 0.000890 0.855522 0.000718 0.860015 0.000879
50 0.756570 0.001241 0.804625 0.000673 0.813316 0.000822
60 0.672137 0.002908 0.746005 0.000909 0.758855 0.001158
70 0.579394 0.003105 0.679557 0.000984 0.691942 0.001030
80 0.368391 0.006010 0.604714 0.001191 0.604779 0.001498
90 0.093756 0.001267 0.510287 0.001127 0.476452 0.002035

(d)

MSSIM results: Lenna
Noise % SVM-M1 SVM-M2 SVM-R

10 0.985231 0.000228 0.985689 0.000188 0.985842 0.000181
20 0.964831 0.000407 0.969028 0.000366 0.969652 0.000358
30 0.936021 0.000549 0.949325 0.000391 0.950161 0.000462
40 0.891417 0.003147 0.925923 0.000477 0.926024 0.000398
50 0.847043 0.001220 0.896965 0.000344 0.893932 0.000713
60 0.777944 0.002245 0.862579 0.000881 0.850796 0.001099
70 0.679344 0.002105 0.820914 0.000564 0.788319 0.001127
80 0.386960 0.004851 0.768412 0.000831 0.693430 0.001324
90 0.046905 0.000822 0.687110 0.000864 0.526401 0.002583