A CGA-MRF Hybrid Method for Iris Texture Analysis and Modeling

Abstract

This paper proposes a novel framework for iris image processing based on conformal geometric algebra (CGA) and Markov random field (MRF). Texture complexity and individual differences are two unique features of iris image, which bring many difficulties to automatic analysis and diagnosis. We propose a circle detection algorithm based on CGA for iris image segmentation. The algorithm is simple and has a wide scope of application. What's more, it can detect the inside and outside boundaries of iris simultaneously without any denoising. Then we propose a novel scheme for texture representation of iris image based on MRF. By learning the statistical texture differences of different pathological features, such as holes, cracks, a MRF based texture representation method shows different pathological regions in iris. Experimental results demonstrated that the proposed framework is very practical, provides a great help for subsequent diagnosis as well.