A fast method for automated detection of blood vessels in retinal images

This paper presents a fast method for automated detection of blood vessels in retinal images. Edges of blood vessels are first extracted by Sobel operators. To make the algorithm fast, edge thinning is employed to preserve only one seed point along each cross section of the vessels. Local windowing and local thresholding are then applied to each preserved seed point to transform the local gray-scale images to a binary map. Twelve binary matched filters are employed to search for vessel segments and the final vessel map is obtained by a single-linkage region growing technique with adaptive thresholds.