Illumination Invariant Feature Extraction for Multispectral Palmprint Verification

Abstract

The aim of biometrics is to identify humans from their personal traits more efficiently using devices, algorithms and procedures for applications that require security and authentication. Multispectral image analysis has gained importance due to its potential for accurate and faster recognition performance. In this paper, Multispectral palmprint biometric system is proposed which uses the fusion of both MS and visible image to acquire more discriminative palm print information. The proposed system collects palm print images in visible and NIR bands. PCA based Fusion algorithm has been used to obtain more informative palmprint. First, Region of Interest (ROI) is extracted from the acquired palm print images. Then, features are extracted using phase congruency, histogram of gradient, Gabor filter and adaptive thresholding based algorithms. Simple distortion based measures are used for recognition. The proposed system is tested on a palmprint data collected using 080GE multispectral camera. Simulation results show high recognition performance using Gabor features obtained by fusion of visible and NIR palm print image.