Abstract: Hard exudation is one of obvious symptoms in diabetic retinopathy.The bright speckles in SD-OCT(Spectral Domain Optical Coherence Tomography)have close relation with exudation.In order to research the relation between retinopathy and exudation,it is necessary to extract the bright speckles.However,there are few studies about the segmentation of bright speckles.In this paper,we first limited target regions by layer segmentation methods,then determined the seeds sets by self-adaption threshold and finally extracted bright speckles by region growing based on human vision feature.Experiments demonstrate that our method can accurately segment the bright speckles in diabetic retinal images.

Key words: Spectral domain optical coherence tomography,Diabetic retina,Hard exudation,Layer segmentation,Self-adaption threshold,Region growing