A method of multispectral finger-vein image fusion

The optical characteristics of the biological tissue varies with the spectral of near infrared light, so the multi-spectral finger-vein images will present some different tissue information compared with the single-spectral finger-vein images. Therefore, the fusion of multispectral finger-vein image will be more beneficial for finger-vein recognition. In this paper, we propose a pixel-level based fusion method which utilizes an improved regional energy fusion strategy for implementing multispectral finger-vein fusion. The proposed fusion method takes the regional energy in a 3*3 sliding window as fusion intensity, and for dealing with the redundant information in some bands, the optimal band combination is adopted in fusion process. Moreover, the proposed strategy only cost a few milliseconds in computing, which can meet the real-time requirement. Finally, experimental results show that the proposed method achieves an excellent fusion of finger-vein multispectral information and provides more accuracy in finger-vein recognition.