Using the input image (where the quality score, i.e., the FVQ derived using Eq.
The learning database of 17 people was only used to identify the optimal threshold for FVQ. During testing, the finger-vein images of only 16 people were used for enrollment and recognition (i.e., it did not use the images from the other 17 people).
Therefore, we compared the accuracies of the two schemes using SUM and AND rules to consider the quality scores (FVQ using Eq.
With the AND rules, both the enrolled and input images were good-quality images because the following conditions were satisfied: "the quality score (FVQ in Eq.