Kernel approach for similarity measure in latent fingerprint recognition

The Recognition of Fingerprint is one of the fundamental problems in the field of pattern recognition. Unfortunately, accuracy of Latent fingerprint matching is still difficult implication and challenging until today. To find the similarity between two images is a trivial task. This process becomes more challenging and risky, when among two input images one is poor quality, such as a latent fingerprint. Latent fingerprints are distorted, partial and having background noise. The main motto of this research is to design an intelligent procedure equivalent to human perception in matching the latent to exemplar fingerprint scenario. In this paper, a new Kernel-based structural similarity measure algorithm is designed for match score computation. The proposed approach is more robust to invariance such as scale change and rotation in the input image. The result describe, that the similarity score value is improved by 1.6% on an average as compared to existing similarity calculation approach.