Automatic Generation of Synthetic Palm Vein Images: a Nature-based Approach

Biometric systems are essential in modern society because of the growing demand for security. Palm vein patterns have become an emerging biometric trait during the last decade. Despite that, the publicly available datasets for evaluating palm vein-based biometric approaches are characterized by a limited number of subjects. Therefore, this work aims to propose a method for the generation of synthetic palm vein images by simulating the vascular structure of the vein plexus in the palm. The proposed approach is inspired by the growth of the Physarum polycephalum, which is an optimized growth system based on natural selection by removing unnecessary pathways from a randomly generated natural-based transport network. The generated vascular network is enhanced by adding palmprint details in order to obtain more realistic images. We have obtained a database comprised of 2,000 different subjects with six samples per each, which is 8 times bigger than the largest publicly dataset. Experimental results show that synthetic images generated by our model share visual characteristics with real images. Finally, the generated dataset, called Natural-based Synthetic Palm Vein Database (NS-PVDB), is publicly available on the website of our laboratory.