Accurate Iris Location Based on Region of Interest

Iris location is an essential module in iris recognition. Traditional iris location methods involve a large range of search, which is computation wasting and sensitive to noise. And these methods adopt circular template to locate the pupillary boundary. It may not accurately describe the pupillary actual boundary and bring the error for the following feature extraction and recognition. To address these problems, this paper presents an algorithm of accurate iris location based on region of interest for improving the accuracy of iris location. At first, according to the feature of approximate concentric circles of iris inner and outer boundaries, the Region of Interest (ROI) only containing the complete iris information can be automatically extracted from an original iris image by Histograms of Oriented Gradients (HOG) to get the statistical information of direction and gradient of an iris image and then the information is taken into Support Vector Machines (SVM) for training and the SVM decision function is gotten. It can eliminate the sensitive noise and reduce the amount of calculation when reserving the useful information as much as possible. Then the iris inner boundary is located roughly by using minimum average gray method, on the basis of this, the annular region is mapped on the rectangular region for the iris inner boundary accurate detection. At last, the iris outer boundary is confirmed by using the improved J.Daugman circle differential algorithm. Experimental results show that the proposed algorithm can efficiently improve the accuracy of iris location.