Finger-Vein Recognition Based on Fusion of Pixel Level Feature and Super-Pixel Level Feature

Abstract

Finger-vein is a promising biometric technique for the identity authentication. However, the finger displacement or the illumination variation in image capturing may cause bad recognition performance. To overcome these limitations, multi-biometric system, an effective method to improve the performance, is proposed. In this paper, a new multimodal biometric system based on pixel level feature and super-pixel level feature is proposed. First, the pixel level feature and the super-pixel level feature are extracted and matched by the Euclidean distance respectively. Then, pixel-super-pixel fusing score (PSPFS) is generated by the weighted fusion strategy. At last, the PSPFS is used to make the decision. Experimental results show that the proposed fusion method not only has better performance than the methods using single level feature, but also outperforms the fusion methods based on the fusion of two pixel level features.