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

Table 4

Table results in mean MSSIM for different impulse noise reduction methods. The mean is obtained for different noise ratios in 10 test images. Total column reflects the total mean for all ratios and images.

Method	Noise percentage
	10	20	30	40	50	60	70	80	90	Total
ACWM [19]	0.930	0.876	0.734	0.502	0.277	0.141	0.069	0.033	0.014	0.397
AMF [12]	0.922	0.909	0.884	0.847	0.800	0.742	0.653	0.466	0.139	0.707
DBAIN [20]	0.978	0.951	0.918	0.877	0.825	0.761	0.677	0.570	0.427	0.776
DSFIRE [21]	0.961	0.905	0.783	0.575	0.348	0.183	0.088	0.040	0.016	0.433
FIDRM [22]	0.977	0.948	0.912	0.870	0.821	0.767	0.696	0.597	0.456	0.783
MMEM [14]	0.943	0.918	0.893	0.868	0.837	0.801	0.754	0.693	0.596	0.811
NAFSM [23]	0.968	0.938	0.904	0.868	0.828	0.784	0.732	0.665	0.538	0.803
PSMF [17]	0.856	0.836	0.802	0.754	0.671	0.509	0.272	0.045	0.019	0.529
SDROM [5]	0.925	0.883	0.829	0.754	0.645	0.508	0.343	0.179	0.067	0.570
MDBUTMF [24]	0.979	0.956	0.930	0.897	0.851	0.765	0.608	0.384	0.178	0.728
SVM-M1 (32-40)	0.978	0.950	0.912	0.859	0.802	0.723	0.617	0.366	0.077	0.698
SVM-M1 (64-40)	0.978	0.950	0.912	0.860	0.803	0.725	0.619	0.369	0.077	0.699
SVM-M2 (32-40)	0.978	0.954	0.926	0.893	0.853	0.807	0.752	0.685	0.585	0.826
SVM-M2 (64-40)	0.978	0.954	0.925	0.892	0.852	0.806	0.751	0.684	0.584	0.825
SVM-R (32-40)	0.976	0.953	0.926	0.894	0.854	0.803	0.735	0.636	0.478	0.806
SVM-R (64-40)	0.976	0.953	0.926	0.894	0.854	0.803	0.735	0.636	0.479	0.806