Alignment Free Partial Biometric Recognition for Fingerprints and Face

Abstract

Biometric recognition with holistic image approaches has already attained much satisfactory results. In the past decade, research attempts were mainly focussed on matching partial biometric traits. The partial nature of these traits is affected more in finger and face images that are easily subjected to occlusions. Images in real time situations are more often of these kinds and can significantly deteriorate the performance of the recognition system. With regard to overcome the defiance in partial biometric recognition, an efficient wavelet based SIFT algorithm is proposed. Biorthogonal wavelet basis (4.4) is considered for obtaining the discrete wavelet transform of the finger and face images. The Scale Invariant Feature Transform (SIFT) algorithm is exercised only on the selected sub bands (LL and HH). The partial biometric database is procured by concealing the entire image with various degrees of orientation (vertical, horizontal, diagonal and random). In the analysis, occlusions at different rates such as 15%, 30%, 45%, and 60% of the holistic image are regarded. Experimental studies using finger and face images with 100 individuals reveal that the reported approach reduces False Rejection Rate (FRR), False Acceptance Rate (FAR) and it improves the recognition accuracy.