A unified approach for automated segmentation of pupil and iris in on-axis images

We propose a unified approach for the automatic and accurate segmentation of the pupil and iris from on-axis grayscale eye images. The segmentation of pupil and iris is achieved with Basis-spline-based active contour and circular active contour, respectively. The circular active contour shape template has three free parameters, i.e., a pair of center coordinates and the radius. Basis-spline has M knots in the shape template and five free parameters i.e., a pair of center coordinates, scaling in the horizontal and vertical directions, and the rotation angle. The segmentation of the region of interest is done by minimization of the local energy function. Optimization of the local energy function of circular and Basis-spline-based active contour is carried out using gradient descent technique and Green’s theorem. To achieve the segmentation of iris boundary, the circular active contour method is combined with our novel occlusion removal algorithm. This helps in removing eyelid and eyelash occlusions for accurate iris segmentation. Automatic localization of the pupil is achieved by the sum of absolute difference method. The proposed algorithm is validated on three publicly available databases: IIT Delhi Iris, CASIA Iris Interval V3, and CASIA Iris Interval V4 databases consisting of 7518 grayscale iris images in total. For the segmentation of pupil from the aforementioned databases, we attained a Dice index of 0.971, 0.950, and 0.960, respectively, and for the segmentation of iris, we attained a Dice index of 0.905, 0.898, and 0.900, respectively. An exploratory data analysis was then done to visualize the distribution of the performance parameters throughout the databases. The segmentation performance of the proposed algorithm is on par with that of the reported state-of-the-art algorithms.