Automatic segmentation of optic disc using modified multi-level thresholding

Diabetic retinopathy and glaucoma are chronic eye diseases that lead to blindness if not diagnosed and treated in time. In the study of the different stages of Diabetic Retinopathy (DR) and Glaucoma, the use of retinal image analysis has become indispensable. The structure of the optic disc, the cup-to-disc ratio (CDR) and the thickness of the blood vessels are indicators of the condition of the eye. This paper proposes a novel method of automated OD (optic disc) segmentation using a modified multi-thresholding technique on a pre-processed fundus image. The algorithm we propose proceeds through three main steps: Pre-processing on the fundus image, followed by segmentation using the proposed novel multi-level Thresholding method and finally using morphological operations over the processed image to obtain the segmented results. The results from this method of segmentation are compared with those obtained using K-means clustering algorithm and the superiority of our method is demonstrated.