Robust Orientation Field Estimation and Extrapolation Using Semilocal Line Sensors

Abstract

Orientation field ( OF) estimation is a crucial preprocessing step in fingerprint image processing. In this paper, we present a novel method for OF estimation that uses traced ridge and valley lines. This approach provides robustness against disturbances caused, e. g., by scars, contamination, moisture, or dryness of the finger. It considers pieces of flow information from a larger region and makes good use of fingerprint inherent properties like continuity of ridge flow perpendicular to the flow. The performance of the line-sensor method is compared with the gradients-based method and a multiscale directional operator. Its robustness is tested in experiments with simulated scar noise which is drawn on top of good quality fingerprint images from the FVC2000 and FVC2002 databases. Finally, the effectiveness of the line-sensor-based approach is demonstrated on 60 naturally poor quality fingerprint images from the FVC2004 database. All orientations marked by a human expert are made available at the journal's and the authors' website for comparative tests.