Hemorrhage segmentation using mathematical morphology and digital image processing

Abstract

Diabetic Retinopathy is one of the complication diseases caused by diabetes mellitus. This disease attacks the retinal eye of human. There are three stages of this disease (mild, moderate and severe). Appearance of Hemorrhage is one of any symptoms included in Non Proliferative Diabetic Retinopathy. Hemorrhage appears on the second and third stages. This study is aimed to detect existences of hemorrhages. We employ morphological operations, filtering techniques, removing blood vessels and macula to detect them. This segmentation model achieves the best accuracy of 85.25% for the retinal images.