An NMF-based method for the fingerprint orientation field estimation

Abstract

Fingerprint orientation field estimation is an important processing step in a fingerprint identification system. Orientation field shows a fingerprint's whole pattern and globally depicts the basic shape, structure and direction. Therefore, how to exactly estimate the orientation is important. Generally, the orientation images are computed by gradient-based approach, and then smoothed by other algorithms. In this paper we propose a new method, which is based on nonnegative matrix factorization (NMF) algorithm, to initialize the fingerprint orientation field instead of the gradient-based approach. Experiments on small blocks of fingerprints prove that the proposed algorithm is feasible. Experiments on fingerprint database show that the algorithm has a better performance than gradient-based approach does. © 2012 Springer-Verlag Berlin Heidelberg.