Sliding Box Method for Automated Detection of the Optic Disc and Macula in Retinal Images

Abstract

In this paper we propose two simple and efficient algorithms for automated detection and localization of optic disc and macula with high accuracy. In the learning phase a set of statistical and fractal features were tested on 40 images from STARE database. The selected features combine spatial and spectral properties of the retinal images and are based on minimum or maximum criteria. In the first step of the algorithm, a sliding box method is used for primary detection of optic disc and macula. For the second phase, a non overlapping box method is proposed for accurate localization of the center of the optic disc. The features are different for the two phases and also for the two cases optic disc and macula. The algorithms were tested on a set of 100 retinal images from the same database. By accurate determination position of the optic disc and macula, the results confirm the efficiency of the proposed method.