Human identification from at-a-distance images by simultaneously exploiting iris and periocular features

Iris recognition from at-a-distance face images has high applications in wide range of applications such as remote surveillance and for civilian identification. This paper presents a completely automated joint iris and periocular recognition approach from the face images acquired at-a-distance. Each of the acquired face images are used to detect and segment periocular images which are then employed for the iris segmentation. We employ complex texture descriptors using Leung-Mallik filters which can acquire multiple periocular features for more accurate recognition. Experimental results presented in this paper achieve 8.1% improvement in recognition accuracy over the best performing approach among SIFT, LBP and HoG presented in the literature. The combination of simultaneously segmented iris and periocular images achieves average rank-one recognition accuracy of 84.5%, i.e., an improvement of 52% than those from only using iris features, on independent test images from 131 subjects. In order to ensure the repeatability of the experiments, the CASIA.v4-distance, i.e., a publicly available database was employed and all the 142 subjects/images were considered in this work.