Skip to Main Content
A fast, reliable and efficient method for detecting the optic disc and exudates in retinal fundus images is presented in this work. The algorithm proceeds through three main steps: 1. Segmentation of blood vessels using median filtering and morphological operations and detection of the convergent point by fitting the blood vessels data using least square polynomial curve fitting algorithm. 2. Extraction of brighter regions using multilevel thresholding that includes optic disc and exudates. 3. Determination of optic disc (OD) among brighter regions from the fact that the convergent point (CP) of blood vessels falls within optic disc and label other brighter regions as exudates. The method has been tested on normal, abnormal retinal images. The Accuracy (M), Sensitivity (S) and Predictive value (PV) are also presented and found this algorithm provides promising results.
Date of Conference: 4-7 Jan. 2005