Automatic Fovea Localization in Fundus Images

摘要: One of the most common modalities to examine eye- background is fundus image. These images are photographs taken through pupil by a camera. The evaluation these im- ages usually carried out ophthalmologists or experts during visual inspection. aim our work accelerate this process fully- automatic screening. To enable disease and tissue specific feature extraction, it necessary segment different visible struc- tures in images. Our group already published methods extract vessel tree estimate position diameter optic nerve head. In paper, we present localize fovea macula region, compare other approaches using High Resolution Fundus database. showed, that proposed method capable localizing region with an average dis- tance error 0.1 times disk diameter.