Automated classification of contact lens type in iris images

Abstract

Textured cosmetic lenses have long been known to present a problem for iris recognition. It was once believed that clear, soft contact lenses did not impact iris recognition accuracy. However, it has recently been shown that persons wearing clear, soft contact lenses experience an increased false non-match rate relative to persons not wearing contact lenses. Iris recognition systems need the ability to automatically determine if a person is (a) wearing no contact lens, (b) wearing a clear prescription lens, or (c), wearing a textured cosmetic lens. This work presents results of the first attempt that we are aware of to solve this three-class classification problem. Results show that it is possible to identify with high accuracy (96.5%) the images in which a textured cosmetic contact lens is present, but that correctly distinguishing between no lenses and soft lenses is a challenging problem.