Segmentation of blood vessels in fundus color images by Radon transform and morphological reconstruction

In this paper we have presented an algorithm for vessel detection in retinal color images which works based on local Radon transform and morphological reconstruction. In our previous work, the algorithm was applied to conjunctival images and the performance was evaluated subjectively. In this work an extended version of the algorithm is applied to retinal images and our aim is to detect vessels in retinal images of patient people and make a good discrimination between blood vessels and other pathological elements like microaneurysms and hemorrhages which have the same color as vessels. The algorithm performance was evaluated numerically by applying it to DRIVE image database and the achieved ROC curve demonstrates its high accuracy.