Enhancement of latent fingerprints on banknotes using monogenic wavelets

Latent fingerprints on banknotes provide valuable information in forensic applications to identify suspects of robbery cases and counterfeiters. The image quality of latent fingerprints is poor due to the complex background of banknotes compared to live-scan images acquired in a controlled environment. An effective orientation field estimation algorithm is necessary for enhancing poor quality latents in the presence of background noise for reliable extraction of minutia features. There have been attempts in the literature to characterize a fingerprint using an AM-FM model. Motivated by the ability of the monogenic signal representation to demodulate such a signal to extract its features, this paper proposes a novel algorithm of latent fingerprint multiresolution orientation estimation with a modified Simoncelli isotropic wavelet design and by an appropriate use of the multiresolution orientation field. Experimental results on synthetic noisy images and real latents from our created latent fingerprint database on banknotes show that the enhanced image quality by using the proposed algorithm is better than the other existing methods for estimating the fingerprint orientation.