The Personal Recognition Algorithm Study Based On Iris

Iris Recognition as one kind of biometrics develops very fast in recent years, the study on this technology has significant value both theoretically and practically, the prospect is promising. On the basis of analyzing existing recognition methods, we mainly make some improvements and innovation on iris image preprocessing and feature extraction.In iris localization, a method which uses binarizing and Freeman chain code is proposed to localize the inner border of iris, based on transcendental information, the outer border is detected using Canny operator and circular Hough transform.In order to exclude the disturbing information, this paper propose an algorithm which combines closing operation and linear Hough transform to detect eyelid, in this way, the eyelash's adverse impact on eyelid localization can be effectively minished; based on Kong and Zhang's method, We use 1-D Gabor filter and medfiltering to detect eyelash; mirror reflection is eliminated by means of thresholding. The orbicular iris is unwrapped to a 40×480 rectangular block using polar coordinate transform. By improving the method of Ma's histogram equalization, the normalized image is enhanced.A feature extraction algorithm based on segmentation of Log-Gabor subbands is proposed in this paper. This method uses 2-D Log-Gabor filters to extract the textural feature of iris, and feature codes are created through segmenting the subbands of middle and high frequency into several blocks and coding the phase of every block's maximum point in succession. When doing the matching, hamming distance is used as the measurement of similarity between irises.At last, for validating the performance of algorithm, we do the experiments on the CASIA database.