Iris Localization Algorithms Based On Radial Symmetry Transformation

As the loading subject of applied mathematics, pattern recognition and image processing and other interdisciplinary topics, iris recognition technology has been given more and more attention. In recent years, more and more people dedicate to the study of iris recognition. however, because the conditions during iris images acquisition is subjected to restrictions, pictures'quality is difficult to ensure, this brings a certain degree of difficulty to iris recognition. So, With Matlab as the tool, main works about iris location in iris recognition system are described following.(1) The iris recognition principle and process were introduced, and the current domestic and foreign classical design methods of iris recognition system, which has four parts:image acquisition, image preprocessing, feature extraction, feature matching, were introduced. In addition, the actuality of iris location algorithms was introduced briefly.(2) In this paper, the images we used are from CASIA-Iris-Thousand database, and we proposed an iris location algorithm based on the radial symmetry transform. First of all, on small-scale picture, detecting light spot position and eliminating the interference, then ensure the center we find is inside the pupil, simultaneously, make sure that the radius is with little deviation, in this premise, we locate the center and radius of the pupil roughly. After those, cutting effective pupil region from the original image. Then, we apply the radial symmetry transform once again to detect the center and radius on the image captured. Because of radial symmetry transform is only for symmetric image, then after it can be quickly and accurately positioning the center and radius of iris on the image captured.(3) On the base of iris localization, we train a large number of iris pictures, then receive a eyelash occlusions curve. By this curve we can detect the eyelash occlusions area effectively, ruled out the interference of eyelash finally.Our conclusion is based on the images in CASIA-Iris-Thousand database, the experimental result achieved 100 percent.