Abstract: Corresponding to the problem of difficult feature extraction , poor stability and large characteristic dimension of palmprint, a novel palmprint recognition algorithm based on Curveletinvariant features of the surface is proposed.We can obtain the stable feature surface though Curvelet transform,which is used to match.This way not only simplifies the operation, such asfeature extraction, image coding and other traditional operation,but also has lower dimension, which leads tohigh stability and fast recognition speed with a high recognition accuracy at the same time. Finally, we use the normalized correlation classifier to measure the similarity. By PolyU palmprint database verification, the equal error rate of our algorithmis only 1.7690% and matching time is 16.6 ms, which demonstrates the effectiveness of the proposed algorithm.

Key words: palmprint recognition, Curvelet, invariant features of the surface, normalized correlation classifier, FAR and FRR curves