A fully-automated retinal blood vessels detection using filling algorithm

Abstract

This paper proposes a fully-automated retinal blood vessels detection method using image preprocessing, multi-scale line detector and post-processing approaches. In this method, a filling algorithm applied on the green channel has solved the vessel problem with central reflex. Moreover, the sum of multi-directional top-hat transformation has successfully enhanced all vessels including thin and tortuous vessels, as well as reduced the optic disk effect in retinal images. The proposed method has shown to be comparable to previous work. In addition, experimental work conducted using HRF images has also shown promising results with average 79.01% (sensitivity), 94.64%(specificity) and 93.30%(accuracy). © 2017 International Information Institute.