Environmentally realistic fingerprint-image generation with evolutionary filter-bank optimization

Constructing a fingerprint database is important to evaluate the performance of an automatic fingerprint recognition system. Because of the difficulty of collecting samples, there are only few benchmark databases available. Moreover, it is hard to evaluate how robust the system is against various environments with those databases. This paper presents a novel method that generates fingerprint images automatically from only a few training samples by using the evolutionary algorithm. Fingerprints generated by the proposed method include similar characteristics of those collected from the corresponding real environment. The proposed method has been verified by comparing with real fingerprints, indicating the usefulness of the method.

► We model environmental effects in the benchmark database for fingerprint recognition.
► It automatically generates fingerprint images through the interaction with experts.
► Evolutionary computation helps to find an optimal combination of simple filters.
► It does not rely on any complicated expert knowledge on image processing.
► It can be applied to various problems of image generation and enhancement.