Noise-resistant network: a deep-learning method for face recognition under noise

EURASIP Journal on Image and Video Processing

Table 5 The average recognition rates of different methods on the AR database injected with salt & pepper noise

Algorithm	Chi-square distance, d =				Histogram intersection, d =				Modified G-statistics, d =
Algorithm	0.05	0.10	0.15	0.25	0.05	0.10	0.15	0.25	0.05	0.10	0.15	0.25
FLBP	0.6959	0.6236	0.4462	0.2794	0.6908	0.6051	0.4277	0.2669	0.6938	0.6103	0.4451	0.2753
NRLBP	0.7149	0.6743	0.5605	0.3127	0.6995	0.6526	0.5663	0.3075	0.7077	0.6774	0.5742	0.3142
NRLBP+	0.7344	0.7014	0.6112	0.3943	0.7138	0.6978	0.6041	0.3822	0.7210	0.7059	0.6089	0.4075
NRLBP++	0.7390	0.7122	0.6242	0.4148	0.7249	0.7088	0.6158	0.4043	0.7365	0.7154	0.6228	0.4297
	Pearson correlation coefficient, d =				Euclidean distance, d =				Cosine distance, d =
	0.05	0.10	0.15	0.25	0.05	0.10	0.15	0.25	0.05	0.10	0.15	0.25
BN2	0.8023	0.7615	0.7002	0.5928	0.8144	0.7502	0.6948	0.5965	0.8014	0.7793	0.7046	0.5943
BN1	0.8310	0.8012	0.7534	0.6443	0.8245	0.7993	0.7315	0.6178	0.8327	0.8089	0.7412	0.6332
NR-Network	0.8542	0.8327	0.7886	0.7013	0.8487	0.8214	0.7749	0.6932	0.8597	0.8412	0.7924	0.7107