Retinal Blood Vessels Segmentation Using Classical Edge Detection Filters and the Watershed

Eye diseases are among the leading health risks people face today, with retinal blood vessel extraction being a crucial step in analyzing retinal images for medical diagnoses. The structures and geometric features of retinal blood vessels offer valuable insights into a patient's health and contribute to the diagnosis of conditions of eye diseases in addition to human body diseases such as diabetes and high blood pressure. The goal of all researchers is to prevent blindness by accurate diagnosis and timely treatment of these diseases. In this paper, we suggested an approach for extracting retinal blood vessels that consists of five stages. The first stage involves preprocessing, where the green channel of the input color image is selected, and edges are detected. In the second stage, we label all connected regions, close gaps, and fill holes. The third stage focuses on improving image contrast, applying a Gaussian filter, and using mathematical morphology to highlight the vessel network. The fourth stage involves segmenting the image into several regions using mathematical morphology and the watershed algorithm. Finally, in the fifth stage, small areas not belonging to blood vessels are removed based on their area and perimeter. Experimental results proved that the suggested approach produces a high-quality map of potential blood vessels. The performance metrics confirm the efficiency and accuracy of our approach.