Robust Finger Vein Identification Base on Discriminant Orientation Feature

As a new biometric feature, finger vein has attracted more attention from researchers. In this paper, we propose a new method to improve the performance of finger vein identification systems. Our proposed method includes the following steps: (1) At first, images of finger veins are cropped to have regions of interest (ROI´s). (2) Then, local invariant orientation features are extracted by using MFRAT which handles the finger vein structure(s), variations of illumination and rotation of ROI. (3) And then, Grid PCA is applied to further remove redundant information and form a discriminant representation which is more suitable for finger vein recognition system. (4) Finally, the enlarging training set (ETS) based matching technique is used to overcome the translations. The experimental results on the public finger vein database (SDUMLA-HMT) demonstrate the effectiveness of the proposed method.