Sift-based image alignment for contactless palmprint verification

Abstract

Contactless palmprint recognition is an effective way to improve the user-friendliness of palmprint recognition technology. The main challenge of contactless palmprint recognition is the intra-class variations caused by contact-less image acquisition. In such occasions, traditional palm-print recognition algorithms which require precise image alignment may not contribute. Aiming at solving this problem, this paper proposes a contactless palmprint recognition method with a precise palmprint image alignment. The original contactless palmprint images are firstly aligned using a projective transformation model that estimated from matched SIFT feature points. From the aligned images, a prominent palmprint feature representation method, the competitive code, is extracted and matched. Finally, matching scores of both SIFT and competitive code are fused to further improve the accuracy. Experiments on a public contactless palmprint database show that after the image alignment, the verification accuracy of competitive code has increased dramatically, and the result is further enhanced by fusing the matching scores of competitive code and SIFT features.