A novel method for low-constrained iris boundary localization

Iris recognition systems are strongly dependent on their segmentation processes, which have traditionally assumed rigid experimental constraints to achieve good performance, but now move towards less constrained environments. This work presents a novel method on iris segmentation that covers the localization of the pupillary and limbic iris boundaries. The method consists of an energy minimization procedure posed as a multilabel one-directional graph, followed by a model fitting process and the use of physiological priors. Accurate segmentations are achieved even in the presence of clutter, lenses, glasses, motion blur, and variable illumination. The contributions of this paper are a fast and reliable method for the accurate localization of the iris boundaries in low-constrained conditions, and a novel database for iris segmentation incorporating challenging iris images, which has been publicly released to the research community. The proposed method has been evaluated over three different databases, showing higher performance in comparison to traditional techniques.