Hand dorsal vein recognition based on hierarchically structured texture and geometry features

In recent years, hand dorsal vein has attracted increasing attentions of researchers in the domain of biometrics. This paper proposes a novel approach for hand dorsal vein identification, making use of both the texture and geometry features. The proposed approach works in a hierarchical way: 1) the coarse step segments the vein region and calculates its skeleton to feed the following operations, and the Energy Cost extracted in the Thinning process (TEC) is also used to reduce a large number of false candidates, greatly improving the efficiency; 2) in the fine step, both texture and geometry clues are represented by Local Binary Patterns (LBP) and the graph composed by the crossing points and endpoints of vein pattern respectively, and the two modalities are finally combined for decision making. The proposed method is evaluated on the NCUT dataset containing 2040 hand dorsal vein images of 102 subjects, and the experimental results clear highlight its effectiveness.