Automated Retrieval of Retinal Blood Vessels in Color Fundus Images

Retinal blood vessels (BVs) are crucial anatomical features of the retina, responsible for delivering oxygen and blood. Segmenting these vessels is vital for retinal imaging examination and diagnosing vascular conditions such as Diabetic Retinopathy (DR) and Diabetic Maculopathy (DM). Manual segmentation is labor-intensive and error-prone, necessitating automated methods to identify abnormal retinal vessels and assess disease severity. This paper introduces a novel method for extracting retinal BVs from color fundus images. Initially, the image is pre-processed using Gaussian, median, average filters, and morphological operations. Subsequently, a binary image is generated using global thresholding and further morphological operations to segment the BVs. The final segmented vessels are superimposed on the source image for visualization. The proposed method is validated using the MESSIDOR database and achieved an accuracy of 96.25%.