Automated human identification using ear imaging

This paper investigates a new approach for the automated human identification using 2D ear imaging. We present a completely automated approach for the robust segmentation of curved region of interest using morphological operators and Fourier descriptors. We also investigate new feature extraction approach for ear identification using localized orientation information and also examine local gray-level phase information using complex Gabor filters. Our investigation develops a computationally attractive and effective alternative to characterize the automatically segmented ear images using a pair of log-Gabor filters. The experimental results achieve average rank-one recognition accuracy of 96.27% and 95.93%, respectively, on the publicly available database of 125 and 221 subjects. Our experimental results from the authentication experiments and false positive identification verses false negative identification also suggest the superiority of the proposed approach over the other popular feature extraction approach considered in this work.

► New ear segmentation approach for automated ear recognition.
► False positive identification vs. false negative identification analysis for ear identification.
► High rank-1 recognition accuracy (over 96% for 125 subjects and over 95% for 221 subjects).
► New publicly available ear image dataset for further research efforts.