Compensation for Non-linear Iris Pattern Deformation based on the Tensile Properties of Iris

In general, iris recognition can be performed to identify users based on the unique iris patterns between their pupils and scleras. The main function of iris is to control the size of the pupil according to the amount of environmental light. Therefore, the iris patterns dilate and contract based on changing lighting conditions. To guarantee robust recognition accuracy in spite of these deformed iris patterns, previous research has used the linear rubber band model and uniform track allocation in iris regions. However, some iris patterns are actually deformed nonlinearly due to the complicated movements of certain iris muscles including the sphincter and the dilator. To overcome these problems, we propose a new method of extracting iris features by nonlinear and dynamic track allocation. This proposed method is based on the nonlinear tensile properties of iris patterns. This paper presents two contributions over previous works. First, we automatically detected the nonlinear positions of the iris patterns in the radial direction when the pupil dilated and contracted with changing illumination conditions. This was possible because we used a template matching process with five iris patches. From the process, we were able to allocate tracks in the iris region nonlinearly and dynamically. We then extracted robust iris features for recognition. Second, we found that the nonlinear tensile properties of the iris patterns differed individually. Based on that, we adopted a user-dependent method of dynamic track allocation, which greatly improved recognition accuracy. Experimental results showed that the accuracy of the proposed method was superior to that of conventional methods which use uniform track allocations based on the linear rubber band model. Key-Words: Iris recognition, Tensile properties of iris patterns, Dynamic track allocation.