Face Spoof Attack Recognition Using Discriminative Image Patches

Table 4

Comparison of the proposed method (with SVM, QDA, Naive-Bayes (NB), and Ensemble based classifiers) on CASIA-FASD database with existing methods in terms of EER (%).

Method	Low quality	Normal quality	High quality	Warped photo attack	Cut photo attack	Video attack	Overall
DoG [22]	13.00	13.00	26.00	16.00	6.00	24.00	17.00
IQA [27]	31.70	22.20	5.69	26.10	18.31	34.41	32.45
LBP + SVM baseline	16.50	17.20	23.00	24.70	16.70	27.00	24.80
Multi-LBP baseline	12.77	16.66	26.66	15.55	25.55	17.77	17.77
HOOF [28]	16.66	30.00	26.11	15.55	17.77	38.88	21.11
Mag-HOOF	17.22	33.33	22.77	12.22	20.00	36.60	22.22
HOOF + Multi-LBP	9.44	20.55	16.66	10.00	16.66	24.44	15.55
Motion-MLBP [23]	7.22	13.33	29.44	14.44	22.22	13.33	15.74
Motion magnification [23]	6.11	23.33	13.88	10.00	14.44	20.00	14.44
Color texture [29]	7.80	10.10	6.40	7.50	5.40	8.10	6.20
Proposed: DF-SVM	7.53	6.65	6.28	6.94	8.21	7.97	6.71
Proposed: DF-NB	7.77	7.79	6.66	7.00	7.66	8.16	9.00
Proposed: DF-QDA	5.78	7.01	5.65	6.97	7.88	7.15	7.81
Proposed: DF-Ensemble	4.65	5.99	6.57	5.94	6.49	6.00	6.11
Proposed: CS-SVM	6.46	6.43	5.97	6.56	8.72	7.27	8.54
Proposed: CS-NB	6.69	6.30	5.99	7.49	8.01	7.68	8.69
Proposed: CS-QDA	7.33	8.89	6.09	7.43	8.21	7.68	7.97
Proposed: CS-Ensemble	6.51	5.59	5.30	5.31	6.13	7.12	7.59
Proposed: DEND-CLUSTERING-SVM	7.39	7.09	5.93	7.35	8.22	8.42	8.07
Proposed: DEND-CLUSTERING-NB	6.98	6.72	7.40	7.65	8.23	8.00	8.45
Proposed: DEND-CLUSTERING-QDA	7.11	6.89	7.90	7.03	6.99	8.89	8.67
Proposed: DEND-CLUSTERING-Ensemble	5.89	6.06	5.58	5.33	5.42	6.02	5.16
Proposed: IQA-SVM	6.32	7.00	5.97	7.73	7.71	7.69	7.33
Proposed: IQA-NB	6.66	7.66	6.32	7.33	8.45	8.00	8.22
Proposed: IQA-QDA	6.35	7.78	8.31	8.78	7.87	8.06	8.69
Proposed: IQA-Ensemble	6.02	6.57	5.76	7.27	6.87	7.27	7.27
Proposed: IPI-SVM	8.34	7.99	8.72	7.89	8.56	8.74	8.36
Proposed: IPI-NB	8.88	7.45	8.35	7.22	8.90	8.12	8.83
Proposed: IPI-QDA	6.01	7.33	6.88	7.33	8.78	9.14	8.57
Proposed: IPI-Ensemble	6.00	6.78	6.10	6.27	7.24	8.00	7.22
Proposed: CP-SVM	8.87	8.23	9.08	7.64	9.37	8.90	9.50
Proposed: CP-NB	9.10	9.07	8.39	9.74	9.85	9.45	9.34
Proposed: CP-QDA	8.67	8.34	9.00	8.78	8.51	8.34	8.59
Proposed: CP-Ensemble	8.01	7.27	7.34	6.82	7.13	8.23	7.60
Proposed: MAXDIST-SVM	7.25	5.76	6.68	9.28	8.28	8.42	8.56
Proposed: MAXDIST-NB	7.37	6.98	7.33	7.33	8.00	8.31	8.00
Proposed: MAXDIST-QDA	7.11	6.78	7.32	8.84	8.02	8.42	8.40
Proposed: MAXDIST-Ensemble	5.26	6.00	5.30	5.78	5.49	5.02	5.07