A Segmentation Technique of Retinal Blood Vessels using Multi-Threshold and Morphological Operations

Abstract

Retinal blood vessels are one of the most significant features in the fundus image of the eye, which plays a crucial role in the early screening of different ocular diseases like glaucoma, diabetic retinopathy, cataract and hypertensive retinopathy. Also in the area of biometric systems, blood vessel structure plays a vital role as retina scan is one of the finest and reliable methods. This paper proposed a segmentation technique which accurately extracts retinal blood vessels. The proposed algorithm conducted in three phases (i) pre -processing of image using AHE, CLAHE and average filtering, (ii) Multi-threshold based novel segmentation technique is used and (iii) post processing is done to remove the imperfections and for this morphological operation is employed. This method efficiently segments the vessels and improves the performance parameters. The accuracy of 95.3% is achieved. Implementation part was done in MATLAB 2015a using a DRIVE database openly available online.