An extended Daugman's algorithm for iris with eye pathology recognition

The use of devices and technologies for the collection and processing of biometric data is continuously growing, in particular for the assessment of personal identity for accessing digital services, information systems, and control. A biometric system must be able to recognize and identify an individual by analyzing people's aspects that are unique, easily captured, permanent, and universal. The goal of this paper is to analyze the limits of biometry, in particular in the case of iris recognition. We take Coloboma, a rare eye pathology that leads to a deformation of the pupil, as a case study and we propose an extension of Daugman's algorithms able to perform biometric iris recognition. The algorithm extension impacts the segmentation phase of Daugman's algorithm, where the pupil is approximated by an ellipse instead of a circumference. Our extension obtains an improvement of 32% on the average error in the segmentation phase on pathological eyes. Utilizing this solution will allow individuals with Coloboma to avoid exclusion when accessing services through iris-based authentication.