A finger-vein extraction algorithm based on local Radon transform

Finger-vein recognition, as a emerging biometric technology, has attracted considerable attention in the biometric authentication field because of its unique advantages. The primary task of finger-vein recognition is extracting vein patterns from an infrared finger-vein image. However, due to the light conditions and tissue structure of fingers, the captured image contains vein patterns as well as the noise and irregular shading which can seriously weaken the capability of a finger-vein recognition system in practical application. To settle this problem, in this study, we propose a robust vein pattern extraction method based on local Radon transform. Firstly, given a pixel, we calculate the values along six directions in the neighborhood of the pixel using local Radon transform. Secondly, on basis of the values along six directions, the primary vein orientation and its normal vector can be determined. Thirdly, to further reduce the impact of noise, we compute the curvature along the normal vector and verify whether the pixel belongs to vein patterns by certain rules. Finally, a local binarization method is employed for extracting vein patterns and matched by the template method. The Experiments demonstrate that the proposed approach are robust and reliable for improving the performance of the finger-vein identification system.