Table 6.
Detection errors (APCER, BPCER, ACER, and HTER) of our proposed method with the CASIA dataset using three types of PAI (unit: %).
| PAI | Wrap-photo Access | Cut-photo Access | Video Display | Overall | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| APCER | BPCER | ACER | APCER | BPCER | ACER | APCER | BPCER | ACER | APCER | BPCER | ACER | HTER | |
| Using CNN Features [27] | 3.975 | 2.770 | 3.373 | 0.643 | 2.770 | 1.7065 | 1.810 | 2.770 | 2.290 | 3.975 | 2.770 | 3.373 | 2.536 |
| Using CNN-RNN Features | 1.531 | 1.385 | 1.458 | 0.331 | 1.385 | 0.858 | 0.831 | 1.385 | 1.108 | 1.531 | 1.385 | 1.458 | 0.954 |
| Using MLBP Features | 9.133 | 10.343 | 9.738 | 10.018 | 10.343 | 10.181 | 9.425 | 10.343 | 9.884 | 10.018 | 10.343 | 10.181 | 9.488 |
| FLF | 3.508 | 0.917 | 2.212 | 0.676 | 0.917 | 0.797 | 1.292 | 0.917 | 1.104 | 3.508 | 0.917 | 2.212 | 1.443 |
| SLF | 1.536 | 1.036 | 1.286 | 0.507 | 1.036 | 0.771 | 0.121 | 1.036 | 0.579 | 1.536 | 1.036 | 1.286 | 0.910 |