Mathematical Problems in Engineering

Research Article

Robust Face Recognition via Block Sparse Bayesian Learning

Table 1

Recognition rate on faces with pixel corruption (%).


Method	Dimension	Percent corrupted (%)
Method	Dimension	0	10	20	30	40	50

NN		36.92	42.42	49.67	46.15	28.35	14.95
		48.79	54.95	60.00	59.34	40.88	20.22
		67.25	75.17	79.56	74.73	58.02	35.39
		87.25	93.19	94.95	92.53	76.48	56.04

SRC		54.51	44.62	50.55	46.59	32.75	21.76
		82.64	61.32	66.59	63.52	49.23	28.79
		98.02	85.06	85.28	83.96	71.87	46.37
		100.00	98.24	98.24	97.14	92.09	73.63

BSCO		87.48	83.52	63.30	39.12	23.30	14.29
		98.02	96.04	90.99	75.60	49.45	30.33
		99.34	99.12	97.58	93.85	82.20	58.68
		100.00	100.00	99.12	98.68	97.14	96.70

BSBL		87.25	85.71	68.79	51.43	30.99	19.56
		94.29	92.97	86.15	72.53	59.12	39.34
		99.56	99.34	97.80	92.31	84.18	67.25
		100.00	100.00	99.78	99.12	97.58	89.01