Skip to main content
Fig. 4 | EURASIP Journal on Information Security

Fig. 4

From: Identification performance of evidential value estimation for ridge-based biometrics

Fig. 4

ROCs (ac), EV distribution according to EVA and experts (df) and ccID results (gi). All results are for the Fusion feature set. The first row (ac) shows the top left corner of the receiver operating characteristics (ROCs) for all capture devices. The colour varies according to the fraction of EV and ccID/ccID as the classification threshold moves along the ROC; the smaller the fraction, the lighter the colour (only applicable to “Scanner, None” and “Phone, None”; the fraction equals 1 across the whole range in all other cases). With Global scaling, the second row (df) shows the EV distribution according to the mark’s distortion class, as computed by EVA (solid) and the experts (dashed). The latter is unaffected by the decision threshold and remains constant. The third row (gi) shows the number of ccID marks classified as EV by EVA (light colour), any additional ones by the experts (dark colour) and those not classified as EV by either (white). The gray line is the threshold corresponding to the EER when the operating point moves along the ROC

Back to article page