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

EURASIP Journal on Image and Video Processing

Table 3 The average recognition rates of different methods on the AR database injected with Gaussian noise

Algorithm	Chi-square distance, σ =				Histogram intersection, σ =				Modified G-statistics, σ =
Algorithm	0.05	0.10	0.15	0.20	0.05	0.10	0.15	0.20	0.05	0.10	0.15	0.20
FLBP	0.7864	0.7228	0.5216	0.4227	0.7992	0.7261	0.5341	0.4249	0.7742	0.7253	0.5089	0.4036
NRLBP	0.8005	0.7333	0.5401	0.4175	0.8010	0.7301	0.5291	0.4205	0.7882	0.7107	0.5275	0.4159
NRLBP+	0.7987	0.7547	0.5874	0.4463	0.8023	0.7354	0.5459	0.4388	0.7909	0.7399.	0.5470	0.4334
NRLBP++	0.8094	0.7651	0.6275	0.5056	0.8174	0.7431	0.5948	0.4946	0.7987	0.7363	0.5695	0.4866
	Pearson correlation coefficient σ =				Euclidean distance, σ =				Cosine distance, σ =
	0.05	0.10	0.15	0.20	0.05	0.10	0.15	0.20	0.05	0.10	0.15	0.20
BN2	0.8234	0.7896	0.6731	0.6100	0.8157	0.7766	0.6679	0.6023	0.8274	0.7832	0.6779	0.6088
BN1	0.8368	0.8340	0.7306	0.6868	0.8350	0.8239	0.7189	0.6788	0.8414	0.8301	0.7258	0.6815
NR-Network	0.8514	0.8458	0.7584	0.7062	0.8509	0.8465	0.7452	0.6924	0.8523	0.8483	0.7595	0.7096