The Localization Algorithms For RGB Iris Image

Iris recognition technology is the cross disciplines of applied mathematics, image processing and pattern recognition. Iris localization refers to separating the iris area from the eye images. The iris feature extraction and matching are related to the accuracy of iris location, so the result of iris location directly affects the accuracy of iris recognition systems and efficiency. This paper mainly studies the localization algorithms of different iris images, which can increase the computational efficiency and stability, while maintaining location accuracy.Firstly, this paper describes the iris recognition system, and then summarizes the existing iris location algorithms and improved methods. After that, analyze the advantages and disadvantages of the algorithms.In this paper, a location algorithm based on integro-differential operator is presented for UBIRIS v2.0. The R layer of color image is directly chosen, and two-dimensional Gabor filters and image gradient are used to detect the reflection area. Then the AdaBoost algorithm is selected to locate the iris area, and the upper and lower eyelids must be located. In the polar coordinates of iris image, the radial gradient of iris boundary is obtained. Based on integro-differential operator, the local extremum is taken to search the outer boundary of iris through iterative step by step. Because the inner boundary of iris is weak, the homomorphic filter is chosen to enhance the iris area, and then locate the inner boundary. Experiments show that the location algorithm is efficient, accurate and stable.The Iris location algorithm based on RANSAC is mainly for the iris which is similar to the ellipse or deformation. After preprocessing, the iris image is segmented by the threshold. Then the maximum connected region is used to extract the pupil area. After locate the upper and lower eyelids, iris boundary points are detected in the polar coordinates of iris image. Finally, the RANSAC algorithm based on least squares fitting is used to detect the iris boundary. At last, propose an iris segmentation method based on cluster analysis. According to the iris gray distribution and the maximum variance cluster analysis, the annular area around the location results is selected to be re-segmented. In the end of this paper, experiments verify the algorithm.