Multi-space random projection of face biometric in the radon domain

Abstract

Biometric authentication cannot replace traditional authentication systems unless biometric data is sufficiently protected during the entire authentication procedure. In this paper a novel method to protect face biometric is presented. The randomized Radon signature (RRS) overcomes the natural limitations of biometrics by providing unique, non-reversible and reissuable templates to replace face images in storage and during processing. The face image is transformed into the Radon space where the signatures are constructed and then projected into a random multi-space. Using the eigenface algorithm for evaluation, the results have shown a 252.78% improvement in the separation of the genuine and impostor distributions leading to an 81.49% reduction in the equal error rate.