Adaptive boosted spectral filtering for progressive fingerprint enhancement

Abstract

Adaptive boosted spectral filtering, a novel fingerprint enhancement algorithm, is based on a progressive enhancement and feedback in a spatial-partitioning, frequency-domain approach. The proposed algorithm applies a Gaussian-matched filter starting from high-quality regions and then iteratively propagating good spectra of enhanced ridges to lower-quality regions. The Gaussian-matched filter does not rely on estimation of contextual information such as local ridge orientation and local ridge frequency. This algorithm can effectively enhance the singular point zone and accumulatively improve very low-quality zones. Compared with various enhancement algorithms and some advanced fingerprint modeling with traditional Gabor enhancement algorithms, the proposed algorithm gives the best average equal error rate in 8 out of 15 fingerprint verification competition databases. The proposed algorithm is very promising for the improvement of fingerprint recognition system accuracy in the near future.