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

Table 5

Comparative restoration results in PSNR (dB) and MAE (the second row) for mixed noise (𝑝=20%sp, 𝜎=10: salt-and-pepper and Gaussian; 𝑝=20%rv, 𝜎=10: random-valued impulse and Gaussian; 𝑝=30%: salt-and-pepper and random-valued impulse noise).

Method	Lena image			Boats image
	𝑝=20%sp, 𝜎=10	𝑝=20%rv, 𝜎=10	𝑝=30%	𝑝=20%sp, 𝜎=10	𝑝=20%rv, 𝜎=10	𝑝=30%
Noisy image	12.36	16.03	13.87	12.25	15.76	13.27
	15.54	10.12	15.86	16.70	11.56	18.59
3 × 3 median filter	27.98	29.38	27.94	26.91	27.70	25.86
	3.021	3.081	1.746	3.505	3.388	2.020
5 × 5 median filter	29.39	29.22	28.92	26.87	26.78	26.23
	2.616	2.867	1.998	3.554	3.607	2.549
Bilateral filter	23.66	25.00	22.77	22.72	23.64	21.33
	7.302	6.107	4.205	6.959	5.754	4.587
SDROM	30.84	30.77	33.13	29.23	29.34	29.91
	2.256	2.894	0.980	2.847	3.001	1.401
ROAD-TRIF	31.34	31.53	32.74	29.49	29.62	29.94
	2.221	2.301	1.201	2.823	2.938	1.336
ROLD-EPR	30.95	31.53	32.89	28.91	29.64	30.12
	2.233	2.300	1.140	2.898	2.911	1.201
RORD-WMF	31.63	31.54	33.53	29.75	29.70	30.31
	2.181	2.298	0.917	2.716	2.908	1.129
DWM	30.92	30.79	32.18	29.03	28.89	29.10
	2.251	2.836	1.268	3.317	3.247	1.354
SBF	31.57	30.97	32.81	29.64	29.13	29.97
	2.208	2.599	1.155	2.797	3.120	1.313
SG-DARD-TRIF	31.93	31.56	33.06	30.24	29.79	30.02
	2.177	2.291	1.004	2.645	2.898	1.293