Palm vein recognition system based on multi-block statistical features encoding by phase response information of nonsubsampled contourlet transform

In this paper, we improve our palm vein recognition system to be based on phase response information of nonsubsampled contourlet transform (NSCT). First, we localise the region of interest (ROI), next, we have divided the ROI into a non-overlapping block and we proposed an encoding method based on extracting phase response information of NSCT coefficients, then XOR pattern is applied to extract invariant from local region of the palm vein to create a palm vein template of 512 bytes. Finally, we have calculated the modified hamming distance between templates to estimate the similarity between two palm veins filtered images. The method is tested on the CASIA Multispectral Palmprint Database. The experimental results illustrate the effectiveness of this coding in two modes of biometric palm vein: 99.90% of rank-one recognition rate and 0.19% of equal error rate in verification.