Improving Sensor Interoperability between Contactless and Contact-Based Fingerprints Using Pose Correction and Unwarping

Table 5

Equal error rates (EERs) are given in percentages for contactless to contact-based template comparison.

Enhancement	All (%)	Thumb (%)	Index (%)	Middle (%)	Ring (%)	Little (%)
B					0.85 0.23
CF		0.47 0.02
CA					0.85 0.50
El	1.35 0.10		0.94 0.16	0.66 0.15
Bi				1.00 0.18		1.77 0.15
PC		0.78 0.40
PC + CF		0.68 0.05
PC + CA		0.52 0.06		0.63 0.10
PC + El		0.41 0.16
PC + Bi		0.57 0.16

All comparisons are made using Kauba et al. [12] enhancement for the contactless images. Depicted are the mean and standard deviation of EERs for 100 random subsamples containing 80% of the template comparison scores. The baseline was calculated for the enhanced images after the preprocessing pipeline without any further postprocessing steps. Results for the pose correction (PC) run were calculated by applying pose correction as a postprocessing step to the results of the baseline (B). For the next runs, additional unwarping was added: circular unwarping with a fixed finger width (CF), circular unwarping with an adaptive finger width (CA), elliptic unwarping with a fixed finger width (El), and finally, bidirectional unwarping with a fixed finger width (Bi). Bold entries highlight the best score for this column and italicized entries the best score for the row. Ties are broken by smaller standard deviations.