Fast Optic Disc Localization Using Viola-Jones Algorithm

Abstract

This paper proposes the implementation of the Viola-Jones algorithm for the Optic Disc (OD) localization in eye fundus images. A robust classifier for the detection of the region of interest (ROI), which contains the OD through the Viola-Jones algorithm was trained, and the center of the ROI for the OD localization was found. A 8776-image dataset for training (7300) and testing (1476) the detector was conformed with images from public databases: DRIVE, DRIONS, ROC, HRF, MESSIDOR, DIARETB0 and DIARETB1. A success rate of 99.6% was obtained for the OD localization in an average time of 474 ms. The Viola-Jones algorithm results a fast, accurate and low-computational cost algorithm for the OD localization and it is applicable in the detection and localization of other structures as the macula and fovea. The results are comparable with those of the state-of-the-art algorithms and, in addition, they could be improved in future works and may be applied in eye computer-aided diagnosis or periodic evaluations of the eyes in real-time mobile applications.