A Deep Learning Approach to Segmentation of Distorted Iris Regions in Head-Mounted Displays
Viktor Varkarakis, Shabab Bazrafkan, Peter Corcoran · 2018
In this paper, we consider the next generation of wearable ARlVR display glasses and the challenges of personal authentication on such devices. The use of iris authentication as a mean of creating a seamless biometric link between the user and his personal data offers a viable approach, but due to the likely location of user-facing cameras there are some challenges in achieving an accurate segmentation of the iris. In this paper, a deep neural network was trained to accurately segment distorted iris regions. An appropriate augmentation method is presented to generate the distorted iris dataset used for training from publicly available frontal iris datasets. The proposed method shows promising results in segmenting off-axis iris images in unconstrained conditions.