Estimation of fingerprint orientation field based on improved structure tensor

Abstract

The estimation of fingerprint orientation field algorithm is a key step in the whole fingerprint identification algorithm. This paper proposes an accurate and rapid method by using connected component analysis structure tensor to estimate fingerprint orientation field. Compared to the original structure tensor method, this method combines connected component analysis map and Gaussian kernal rather than only use Gaussian kernal in the construction of structure tensor. It can preserve the image local structure better and make the result more reliable. The experiments results show that this method is fast, accurate and can easy be used by subsequent algorithm. © 2012 Springer-Verlag GmbH.