An Automatic Retinal Optic Disc Segmentation Algorithm using Differential Windowing in the Polar Coordinate Domain

Abstract

Blindness from glaucoma is irreversible but largely preventable if detected early. Information about optic discs can be used to examine severity of certain diseases in the human retina. A novel optic disc segmentation algorithm in the polar coordinate domain is presented in this paper. Automatic optic disc localisation, left or right eye identification, and optic disc boundary extraction are addressed in this paper. A differential windowing operation in the polar coordinate domain is introduced to extract the optic disc boundary in ocular fundus images. The proposed work provides an accurate method to detect left and right eyes using blood vessel convergence. The method is tested using the available databases and real time images, and the results are reported here. The optic disc localisation and left or right eye identification are obtained with 100% accuracy. The experimental results demonstrate the accuracy of the optic disc boundary extraction.