EXPECTING AND COMPARING PALMPRINT FEATURES USING RADON TRANSFORM AND SPARSE REPRESENTATION

Abstract

Abstract Palmprint is the most widely used biometric identifier for personal identification. However, work on principal lines of palmprint is less due to the similarity of principal lines in different individuals. This paper investigates a new approach utilizing the capabilities of sparse representation and radon transform to improve the recognition and verification problems. Gabor filter is used to highlight the textural features in the palmprint which further enhances the performance. In this work, ease of computation of radon transform is exploited while a solution for sparse representation of radon transform orientation dictionary using convex optimization approach. We perform experiments on two publicly available PolyU and IITD palmprint databases and the results were recorded from both the approaches i.e. Radon transform and sparse representation based on radon transform. A significant improvement is observed in case of sparse representation based on radon transform technique than just applying radon transform, thus confirms the practical importance of this approach in real life palmprint identification problems. Since sparse representation based on radon transform outperforms radon transform technique, this approach is also applied on PolyU Knuckleprint database and the results were