Progressive Focusing Algorithm for Reliable Pose Estimation of Latent Fingerprints

Abstract

A pose of fingerprint composes of a reference point and its corresponding orientation. The reliable pose plays a crucial role in fingerprint alignment and registration. Pose estimation of latent fingerprints is a challenging problem. A few methods based on machine learning were proposed in the past decade. These methods try to predict a pose from corrupted information. In this paper, we propose a systematic feedback approach which can remedy the corrupted information while simultaneously estimates a reliable pose. Without manual segmentation, our fully automatic algorithm is able to progressively locate potential poses, and enhance weak friction ridges that form and support these poses through an iterative feedback. Using the NIST SD27 and MOLF DB4 latent fingerprint databases, our experimental results show that our proposed algorithm outperforms the state-of-the-art and existing commercial products for latent fingerprint pose estimation in terms of precision and identification accuracy.