Palm Vein Recognition Using Directional Features Derived from Local Binary Patterns

Abstract

Vein-based biometrics is a newly developed technology for personal recognition, and it is widely used in practice and intensively studied. This paper proposes a method for palm vein recognition based on the directional information derived from local binary patters. In the proposed method, palm vein images are firstly enhanced using a multi-scale Gaussian matched filter to emphasize vein patterns before feature extraction. After that, local binary patterns are extracted from the enhanced palm vein images. Considering that the direction is the most discriminative feature of veins, the directional information is then computed from the local binary patterns. The computed palm vein features are represented as binary series, therefore, similarities can be computed efficiently by binary operation. Experiments carried out over the near inferred band of the PolyU multispectral database shows the superiority of the proposed method on verification accuracy to some state-of-the-art literatures.