A new retinal image processing method for human identification using radon transform

The blood vessels of retinal image have a unique pattern, from eye to eye and person to person. We have used this trait for designed a new person identification system. This approach focused on blood vessels around the optical disc instead of extracting total retinal blood to optimize the computational cost. At first, optical disc is localized using template matching technique and uses it to rotate the retinal image to reference position. This process compensate the rotation effects which might occur during scanning process then a circular region of interest (ROI) around optical disc is selected. Next, a rotation invariant template is created from each ROI by a polar transformation. In the next stage, vessels from each template are enhanced. Radon transform is used for feature definition in our method. Finally we employ 1D discrete Fourier transform and Euclidian distance for feature matching. The proposed algorithm was tested on a 200 image from DRIVE database. Experimental results on the database demonstrated an average identification rate equal to 100 percent for our identification system.