An Efficient Universal Noise Removal Algorithm Combining Spatial Gradient and Impulse Statistic

Table 2

Comparative restoration results in PSNR (dB) and MAE (the second row) for salt-and-pepper noise.

Method	Lena image				Boats image
	𝑝=20%	𝑝=30%	𝑝=40%	𝑝=50%	𝑝=20%	𝑝=30%	𝑝=40%	𝑝=50%
Noisy image	12.42	10.67	9.400	8.460	12.32	10.52	9.300	8.340
	12.42	18.60	24.67	30.95	13.49	20.51	27.27	33.86
3 × 3 median filter	29.57	23.95	19.06	15.39	27.91	23.11	18.90	15.20
	1.663	2.519	4.725	8.473	2.276	3.136	4.834	8.555
5 × 5 median filter	30.22	29.50	28.12	24.41	27.33	26.68	25.70	23.05
	1.963	2.139	2.423	3.052	3.049	3.220	3.390	4.160
ACWMF	31.09	29.81	28.23	24.23	27.41	26.71	25.65	22.89
	1.205	1.886	2.306	3.207	2.977	3.182	3.313	4.404
SDROM	37.74	35.50	32.45	30.81	33.76	30.57	28.86	26.22
	0.511	0.810	1.130	1.546	0.870	1.577	1.991	2.863
ROAD-TRIF	34.81	31.15	27.41	23.10	32.12	27.96	24.18	21.96
	0.847	1.297	2.617	3.612	1.013	2.548	4.177	4.886
DWM	35.64	33.37	30.65	27.02	32.27	29.93	27.95	25.05
	0.789	1.013	1.553	2.652	0.996	1.732	2.780	3.380
DARD-BF	38.44	35.61	33.22	31.34	34.39	31.67	29.39	27.73
	0.382	0.789	1.013	1.290	0.733	1.334	1.672	2.146
SG-DARD-TRIF	38.52	35.75	33.34	31.52	34.45	31.81	29.81	27.94
	0.369	0.757	0.971	1.103	0.716	1.080	1.478	2.091