Pre-enhancing Non-uniformly Illuminated Fingerprint Images
Norman Haas, Nalini K. Ratha, Rudolf M. Bolle
Auto ID 2002 - Tarrytown, NY, IEEE Conference on Automatic Identification
Advanced Technologies, March 2002
Fingerprint image acquisition is a challenging task, as the three dimensional object is mapped to a two dimensional surface. The problem is further compounded by the dirt, oil, dryness and smudginess, over-inking and pressure applied during acquisition. To improve the overall endto- end performance from images obtained with fingerprint images with non-uniform inking or light, we implemented an image pre-enhancer, which attempts to compensate for the non-uniform illumination characteristics of the scanner. Our algorithm contains some knowledge of the scanner design’s geometry, but also attempts to adapt to effects caused by individual unit variations, in the presence of sampling noise. We give quantitative data on the improvements obtained in terms of a receiver operating characteristics (ROC) curve, to show better match accuracies after the images are processed with the proposed algorithm.