Fast and iterative algorithm for iris detection with orthogonal polynomials transform

Abstract

This paper proposes a novel iterative algorithm for iris localization that accurately extracts outer iris boundary and detects eyelids to compensate accuracy problem caused by the non-standard circle characteristic of the iris. The proposed method localizes the outer boundary using image gradient, generated with modified zero-crossing in the second directional derivatives from orthogonal polynomials transform coefficients. Image gradient is the external force to reposition the points present in the two semi circular rings radiating from pupil center. The semi circular rings are iteratively expanded by modifying the radius from inside toward outside to search for the abrupt gradient changes in order to find the iris contour formed by the image feature. In addition, occlusions of the iris images from eyelids are automatically excluded from the detected iris region. The experimental result shows that the proposed scheme is robust in finding exact noncircular outer boundary and outperforms both in accuracy and speed.