Perceptually adapted method for optic disc detection on retinal fundus images

Abstract

This paper presents a novel technique for the detection of the optic disc (OD) in retinal fundus images. The method exploits the color information of the image with a perception adapted approach. CIE L*a*b* color space along with CIE94 color distance are used to obtain 12 color derivatives for each pixel under study. Based on this information, a classifier assigns a probability value to each pixel in the image, meaning its suitability for being part of the OD border. Looking for the pixels with highest probability values, the method detects the basic points for the OD border that are subsequently connected with the livewire technique. The reliability of the tool has been tested with three different classifiers on 198 images from four public available databases obtaining an average success percentage of 85.48% and a mean distance to the closest point of 2 pixels.