Certain Investigation on Latent Fingerprint Improvement through Multi-Scale Patch Based Sparse Representation

Latent fingerprint recognition plays a key task for recognizing and convicting criminals in enforcement society. Latent fingerprint pictures area unit a lot of typically than not of poor quality with blurred ridge structure and numerous overlapping patterns. Advance are achieved on developing machine-controlled fingerprint identification system, it's still difficult to realize reliable feature extraction and identification for latent fingerprints attributable to the poor image quality. Before feature extraction, fingerprint sweetening is critical to suppress numerous noises, and improve the clarity of ridge structures in latent fingerprints. Actuated by the recent success of distributed illustration in image denoising, this paper proposes a latent fingerprint sweetening formula by combining the overall variation model and multiscale patch-based distributed illustration. First, the overall variation model is applied to decompose the latent fingerprint into cartoon and texture elements. The cartoon part with most of the non fingerprint patterns is removed because the structured noise, whereas the feel part consisting of the weak latent fingerprint is increased within the next stage. Second, we have a tendency to propose a multiscale patch-based distributed illustration technique for the sweetening of the feel part. Dictionaries area unit created with a collection of Dennis Gabor elementary functions to capture the characteristics of fingerprint ridge structure, and multiscale patch-based distributed illustration is iteratively applied to reconstruct high-quality fingerprint image. The planned formula cannot solely take away the overlapping structured noises, however additionally restore and enhance the corrupted ridge structures. Additionally, we have a tendency to gift associate degree automatic technique to phase the foreground of latent image with the distributed coefficients and orientation coherence.