基于curvelet变换和形态学的视网膜血管分割-通信与信息系统专业论文.docxVIP

华中科技大学硕士学位论文 华 中 科 技 大 学 硕 士 学 位 论 文 II II Abstract The retina is the innermost layer of the eye, it is composed of many important physiological structures. Vessels are the most important structures. Many diseases can change the shapes and the structures of the blood vessels, so the analysis for retinal vessels can prevent some disease from occurring and diagnose some early diseases. Conventional manual segmentation for retinal images is a time-consuming and trivial task. Intrinsic characteristics of retinal images make the bloods vessels detection process difficult. There are many good methods to detect blood vessels in retinal image proposed by researchers. But most of them trend to lose thin vessels in the low contrast. So it is urgent to make the automatic quick segmentation for retinal images with computer. . Due to the high ability of the Curvelet transform in representing the edges, we propose a method to enhance the retinal image contrast. The method is that modifications of Curvelet transform coefficients to enhance the retinal image edges. We introduce a nonlinear function to modify Curvelet transform coefficients in such a way that details of the small amplitude are enlarged .Definition of the function parameters based on some statistical features of Curvelet coefficients of the input image. The methods can enhance the retinal image contrast. Afterward, we use modified top-hat to detect edges and morphological operators by reconstruction to eliminate the false-edges not belonging to the vessel tree while trying to preserve the thin vessels unchanged. We use multi-structure elements because the directionality feature of them makes it an effective tool in edge detection. Finally, we use the threshold segmentation based on entropy to detect vessels. The performance of the proposed method is evaluated on the available DRIVE databases, and the method can detect all the big vessels and many small thin vessels, and the runtime of our algorithm is below 50s. So o