Research On Optic Cup Detection In Colour Fundus Images

Digital fundus image can not only be used to diagnose eye diseases, but also canprovide the basis for the treatment of diabetes, hypertension and other systemicdiseases. Image processing techniques is able to identify the various structuredfeatures in fundus images, endorsed by major morphological parameters, assistantdoctor for diagnosis and treatment. Thus, digital fundus image processing plays animportant role in both medical research and medical diagnosis.This paper focuses on optic cup detection in colour fundus images. Small cups,low contrast, uneven brightness, vascular disturbance and lots of other factors make itbeing challenging, so generic methods is not applicable to the identification of theoptic cup. CV model based on level set will be used to get the segmentation of theoptic cup in this paper. This model is based on curve evolution and level set theory,the use of global information of the image, which can effectively deal with complextopology, boundaries blur and discontinuous image, it has the advantage of notsensitive to noise.This paper presents an improved CV model in the segmentation of the optic cup,with better performance on segmentation speed, local edge positioning accuracy andself-adaptability. First, we make the colour image sharper by using grayscaleconversion. Then uses mathematical morphology and Otsu method to obtain the initialcontour, which makes initial contour edges closer to the cup, improves thesegmentation accuracy. In the meantime, prior knowledge of the ocular fundus imageswas applied to CV model in a reasonable and timely manner, that makes the modelmore in line with the form of the optic cup segmentation. We also design a betteriteration termination criterion in order to reduce the number of iterations curveevolution. Because of the cup has the characteristics of a circular or elliptical, finallythis paper use the least squares method based on the ellipse fitting algorithm, make theirregular curve becomes the oval shape.Improved CV model overcomes the interference factors like small contrast,uneven brightness and in particular the effects of vascular occlusion in the eye cupssplit. Experimental results show that the optimized CV model and ellipse fitting algorithm improves the speed of the optic cup segmentation and segmentationaccuracy.