A Robust Algorithm for Colour Iris Segmentation Based on 1-norm Regression

In this paper, we propose a novel algorithm for colour iris segmentation. The algorithm may be divided into the following components: coarse iris localization, limbic boundary segmentation, pupillary boundary segmentation, eyelids fitting, reflection and shadow removal. The key contribution of the proposed algorithm is that we demonstrate the power of sparsity induced by l(1)-norm in overcoming the noise and degradations in colour iris images. We show that limbic and pupillary boundary, as well as eyelids, can be fitted robustly by solving l(1)-norm regression problems. The experimental analysis shows the robustness of the proposed algorithm; comparison with state-of-the-art methods achieves an improved performance.