Color Eye Fund Image Cup Segmentation Method Research

Glaucoma is an eye disease that constitutes the second cause of blindness over the world.The occurrence of glaucoma is generally associated with optic disk and optic cup and optic nerve fiber layer changed.The measurement of the optic cup’s parameters is one of the important indicators to detect glaucoma.So accurate segmentation of optic cup has an important clinical significance for glaucoma’s automatic screening and controlling the progression of disease.This paper proposes an optic cup segmentation method for color retinal fundus images based on multi-feature fusion.Firstly,the multi-directional Gaussian difference filter is used to filter the enhanced and smooth retinal fundus vessels to extract the blood vessels in the region of interest,to prepare for the segmentation of optic disc and optic cup.Secondly,optic disk region is segmented by CV model,and the candidate regions of the optic cup are extracted by using fuzzy c-means clustering(FCM)method based on it.Then based on shape and position features of the optic cup,rough segmentation results are obtained by completion,edge fitting and correction of the optic candidate regions.Finally,the corner points detection method based on k cosine curvature is used to locate the vessel bending points,and based on them the rough segmentation results of the optic cup is revised to get accurate segmentation results.The proposed method based on the optic cup appearance and structural features makes the segmentation of the optic cup more accurate.Besides,the accurate of initial contour of vessel bending points is improved by revising the location and shape of optic cup candidate regions.This method is unsupervised without the training,and tested on GlaucomaRepo fundus image dataset,the experimental results show its robustness and effectiveness.