Pixel-based parallel algorithm for retinal vessel tree segmentation

Abstract

Identification of retinal blood vessels allows carrying out early diagnosis of retina. A lot of algorithms have been developed for the segmentation of the retinal vessel tree. Some of these algorithms are based on neural networks or other supervised techniques and show high accuracy but lack of speed. At the same time, some applications may require high speed of segmentation process and less level of accuracy. Our algorithm is fully parallel and based on computation of image convolution thus allowing fast retinal vessel tree segmentation.