Textured detailed graph model for dorsal hand vein recognition: A holistic approach

Abstract

Holistic- and local-based methods are two-pronged in dorsal hand vein recognition, and the latter ones have become dominant recently due to their advanced performance. In this paper, we propose a novel approach to dorsal hand vein recognition using a global graph model which takes both the texture and shape cues into account. We first extend the basic graph model consisting of the minutiae of the vein network and their connecting lines to a detailed one by increasing the number of vertices, describing the profile of the vein shape more accurately. We then append the holistic texture feature of the patch around each vertex, i.e. its PCA coefficients, to make the representation of the graph model more comprehensively. The above two steps significantly improve the discrimination of the graph model, and it reports the rank-one recognition rate of 98.82% on the NCUT dataset. This holistic result is comparable to the ones of most local based methods, demonstrating its effectiveness. Meanwhile, with local texture cues embedded, e.g. LBP, HOG, and Gabor, it further reaches the state of the art accuracy up to 99.22%, showing its good complementarity to local based methods.