Fingerprint enhancement via sparse representation

Abstract

Although fingerprint recognition has been widely studied for personnel recognition in the past decades, it is still challenging problem to achieve reliable feature extraction and recognition for poor quality fingerprints. Fingerprint enhancement is often incorporated prior to feature extraction to improve the quality of fingerprint and achieve higher recognition accuracy. Motivated by the recent success of sparse representation in image denoising, this paper proposes a fingerprint enhancement method by using sparse representation. First, a set of Gabor basis functions with various orientations and frequencies are used to build a redundant dictionary for fingerprint representation. Then, the fingerprint enhancement problem is modeled as an iterative sparse representation of the local patch fingerprint, which can be solved by L1-norm regularized minimization. Experimental results and comparison on FVC fingerprint databases are presented to show the effectiveness of the proposed method on fingerprint enhancement, especially for poor quality fingerprints. © Springer International Publishing 2013.