Palmprint recognition using geometrical and statistical constraints

Abstract

This paper proposes an efficient biometrics system based on palmprint. Palmprint ROI is transformed using proposed local edge pattern (LEP). Corner like features are extracted from the enhanced palmprint images as they are stable and highly discriminative. It has also proposed a distance measure that uses some geometrical and statistical constraints to track corner feature points between two palmprint ROI’s. The performance of the proposed system is tested on publicly available PolyU database consisting of 7,752 and CASIA database consisting of 5,239 hand images. The feature extraction as well as matching capabilities of the proposed system are optimized and it is found to perform with CRR of 99.97% with ERR of 0.66% for PolyU and CRR of 100% with ERR of 0.24% on CASIA databases respectively.