Unsupervised Morphological Approach for Retinal Vessel Segmentation

Glaucoma, diabetic retinopathy, atherosclerosis, hypertensive retinopathy, age-related macular degeneration (AMD), retinopathy of prematurity (ROP) are some of the retinal diseases which may lead to blindness manifest as artifacts in the retinal images. For the early diagnosis of these systemic diseases, retinal vessel segmentation of retinal is initial assignment. In present scenario, retinal blood vessel segmentation automatically and accurately remains as a challenging task in computer-aided analysis of fundus images. In this paper, we put forward an unsupervised morphological approach for automatically extracting blood vessels from retinal fundus images that can be used in the computer-based analysis. Proposed method uses mathematical morphology with modified top-hat transform for preprocessing and hysteresis thresholding for the segmentation of blood vessels. The proposed approach was evaluated on the DRIVE database and is compared with the recent approaches. Proposed method achieved an average accuracy of 95.95% and best accuracy of 97.01% shows that the approach is efficient.