LAUSR

Automatic localization of optic disk is a challenging problem due to the interference of multiple factors. In this paper, we aim at the study of the optic disk localization method and propose a simple and efficient method based on texture orientation voting. This method uses multi-directional Gabor filters to detect the texture orientations of fundus images. Before orientation voting, texture orientation extraction and orientation confidence computation are conducted. Only the texture orientations with high confidence could participate in orientation voting. If the texture orientation of a voter pixel points at a receiver pixel, the receiver will get a vote from the voter, which is weighted according to the positional relationship of the voter and the receiver. The pixel with the maximum vote over the image will be taken as the optic disk location. In a variety of difficult environments, an average OD detection accuracy of 96.8% is obtained and the experimental results verify the efficiency and robustness of our proposed method.