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

Table 3

Table results in mean PSNR 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]	31.24	27.49	22.82	18.48	14.89	12.10	9.78	7.94	6.47	16.80
AMF [12]	32.17	30.45	28.77	27.18	25.59	23.97	21.87	18.03	12.03	24.45
DBAIN [20]	36.58	32.97	30.51	28.50	26.61	24.80	22.90	20.85	18.05	26.86
DSFIRE [21]	34.14	30.13	25.69	21.10	17.14	13.93	11.30	9.23	7.60	18.92
FIDRM [22]	36.28	32.53	30.12	28.29	26.76	25.39	23.89	22.08	19.68	27.23
MMEM [14]	27.77	30.46	29.52	28.71	27.84	26.88	25.76	24.45	22.59	27.11
NAFSM [23]	34.26	31.31	29.54	28.17	27.05	26.03	25.01	23.81	21.22	27.38
PSMF [17]	26.35	25.84	25.08	23.97	21.95	18.91	14.98	9.67	7.61	19.37
SDROM [5]	30.92	28.93	27.07	25.19	22.97	20.53	17.62	14.34	10.79	22.04
MDBUTMF [24]	37.22	33.86	31.62	29.82	27.94	25.53	22.45	18.83	15.05	26.92
SVM-M1 (32-40)	35.49	31.96	29.32	26.53	24.69	22.33	19.76	13.07	5.96	23.23
SVM-M1 (64-40)	35.57	32.15	29.49	26.87	25.10	22.80	20.12	13.28	5.97	23.48
SVM-M2 (32-40)	36.02	32.96	30.86	29.19	27.74	26.42	25.14	23.80	22.09	28.25
SVM-M2 (64-40)	35.96	32.90	30.81	29.14	27.69	26.37	25.11	23.77	22.06	28.20
SVM-R (32-40)	35.63	32.71	30.76	29.14	27.70	26.25	24.71	22.71	19.63	27.69
SVM-R (64-40)	35.47	32.59	30.68	29.10	27.67	26.23	24.70	22.71	19.63	27.64