Illumination-invariant Ear Authentication

Ear biometric research can be traced back as early as 1949. Off late, more work of ear biometrics has been done mainly focusing within constrained environment. Illumination invariant problem in unconstrained environment remains less explored for ear biometric. In this paper, we propose an illumination-invariant ear authentication approach using biometric authentication methods. Ear regions are firstly segmented from side face images using biased normalized cuts and morphological operations. Shape features are then extracted using Log-Gabor filters where the Hamming distance is finally applied for authentication. A authentication rate of 92.66% is reported for intra-class matching on 50 subjects.