Automatic localization and contour detection of Optic disc

In this paper a method for localizing optic disc in retinal images using KL divergence matching technique followed by main vessel detection is proposed. It is seen that localization of the optic disc and retina vessel segmentation is considered as the first step of retinal disease diagnosis and identification algorithm [2], [3]. In this paper algorithm discussed consist of two stages. The first stage is detection of blood vessels using median filtering for background estimation on green channel followed by morphological operation on retinal images. Structural element (SE) is selected in such a way that only main blood vessels are detected. The second stage involves localization of OD. For localization of OD its property of being most bright portion of retinal image is used and histogram matching approach using KL divergence method followed by main vessel verification is used. To evaluate the proposed algorithm we use DRIVE publicly available eye fundus image databases contains 40 images and it achieves accuracy of 100% in locating a point in OD and also on 20 images of STARE database out of which we localize OD in 18 images accurately.