Multi-Perspective Finger-Vein Biometrics

Abstract

Most finger vein recognition systems use palmar finger images. There is some work on the dorsal view, but the remaining views have not been sufficiently investigated yet. All major public available finger-vein databases contain only images from the palmar view and only one smaller database has images from the dorsal view. We aim to fill this gap and evaluate the performance using other perspectives than dorsal and palmar. Therefore, we established a new finger vein data set that consists of videos showing the vein structure all around the finger. We carried out several experiments utilizing common finger-vein recognition algorithms to quantify the recognition performance of each single projection. We further analyzed if a fusion of different views can improve the recognition performance of the system.