Latent Fingerprint Enhancement via Multiscale Patch based Sparse Representation

� Latent fingerprint enhancement method plays an important role in identifying and convicting the criminals in different security based systems. Latent fingerprints are generally of low qualities and unclear separation of ridges. There are three types of fingerprints namely plain fingerprints, rolled fingerprints and latent fingerprints. There are different systems working on plain and rolled fingerprints for their automated identification. The area of latent fingerprints is still not fully automated due to their poor quality. It is still challenging to achieve reliable extraction and identification system for latent fingerprints. Feature extraction and enhancement of fingerprint are the most important operations needed to detect the uniqueness of the fingerprint. This paper proposes a latent fingerprint enhancement algorithm by combining the total variation model and multi scale patch based sparse representation. Total Variation model applied to decompose the latent fingerprint into cartoon component and texture component. Texture components further undergoes with the process of enhancement. Multi scale patch based sparse representation method used to enhance texture components. Experimental results on NIST SD 27 latent fingerprint database are presented to show the effectiveness of the proposed algorithm.