Image quality and position variability assessment in minutiae-based fingerprint verification

Abstract

The assessment of a complete minutiae-based fingerprint automatic verification system, when degrading variability factors are included, is presented. A new large public fingerprint database, the so-called MCYT Fingerprint Database, is used to evaluate the performance of the system in verification tasks. The design of this database, in terms of controlled variability in fingerprint positioning, has made it possible to determine the system performance when the test or the stored images are subject to changes in fingerprint core placing. Some procedures are proposed to cope with this problem, including a multiple-reference strategy. Human supervision and labelling of the image quality of the acquired fingerprint images has also been accomplished, permitting a precise assessment of the proposed minutiae extraction and pattern matching processes. Results, including enhanced procedures for both position variability control and image quality consideration, are presented in terms of DET plots, leading to highly competitive verification scores in terms of EER.