Finger-vein segmentation based on multi-channel even-symmetric Gabor filters

As a new approach to personal identification, finger-vein recognition is becoming an active topic in biometrics. And exploiting the underlying features related to finger-vein networks has been considered as a reliable way for finger-vein recognition. However, finger-vein network segmentation always is a difficult task due to the low contrast of finger-vein images. This paper focuses on finger-vein enhancement and segmentation based on Gabor filters in the spatial domain. Considering the high randomicity of the finger-vein networks, a bank of even-symmetric Gabor filters with eight orientations is firstly used to exploit vein information in images. Then, image reconstruction is implemented to generate an image containing an integrated finger-vein network. Finally, the finger-vein network is segmented using a proposed threshold image method. Experimental results show that the proposed method is capable of enhancing and segmenting finger-vein networks reliably and effectively.