Latent Fingerprint Enhancement Using Sparse Representation

Abstract

Latent fingerprints are fingerprints which are smudged, overlapped, and distorted. They are poor quality images with unclear ridge structure. Latent fingerprints are seen at crime scene, so it plays a vital role in identifying and declaration of criminal in law enforcement.Therefore it is important to enhance the latent finger prints which are distorted by most of the non-finger print patterns. By enhancing the latent fingerprints it becomes easy to match with the large database of know persons fingerprints.By seeing the advances of sparse representation in image denoising, the work proposes the method of enhancement using sparse representation. In Sparse representation initially total variation model will be used for latent fingerprint image, where image is disintegrated into cartoon and texture components. Cartoon component contains noise, which will be discarded while texture component contains main information. Second sparse enhancement algorithm is implemented on texture component of the image. Experimental results are carried on NIST SD4 to show the results of the proposed algorithms.