彩色眼底图像的血管分割方法研究

发布时间：2018-08-12 08:25

【摘要】：眼底血管网络是人体内能够通过非创直接观察到的比较深层次的微血管。任何系统性和血液性的病变都会导致眼底微血管的特征或形态发生变化。眼底图像血管分割是视网膜图像处理及分析中关键性的步骤,对系统性和血液性疾病的早期预防与诊断有较好的研究意义。视网膜图像的特征较为复杂,眼底血管的自动分割很容易受到外界条件和病变自身的影响,而且在眼底图像中,微小血管和背景的对比度低,增加了眼底血管的分割难度,所以提高眼底血管的分割精度是一项重要的研究课题。本文介绍了眼底血管的研究背景和意义及眼球和视网膜的结构特征,阐述了国内外对眼底图像处理的研究现状,分析了血管图像的性质和研究难点,重点从眼底图像去噪和眼底图像血管分割两方面进行了研究,本文采用DRIVE标准图像库和STARE标准图像库中的彩色眼底图像进行仿真实验,完成的主要研究工作如下:(1)改进了一种结合非局部均值滤波的双边滤波眼底图像去噪方法。对眼底图像去噪进行了研究,深入分析了双边滤波与非局部均值滤波两种去噪方法,并对两种去噪方法的优缺点进行总结,在这基础上改进一种结合非局部均值滤波的双边滤波眼底图像去噪方法,与此同时还采用了积分图运算方法和升余弦函数近似灰度相似性函数分别实现NLMF以及BF的快速运算,仿真实验分析结果表明改进的去噪方法对眼底图像的去噪效果较好,运算耗时较少。通过对眼底血管分割方法的分析与学习,本文研究了一种基于水平集函数的彩色眼底血管分割方法。先应用自适应直方图和二维Gabor变换对眼底图像进行预处理,然后对眼底图像运用可变区域拟合能量定义的水平集理论进行血管分割仿真实验。根据眼底血管特点,改进一种基于广义线性模型的彩色眼底图像眼底血管分割算法。该算法先应用自适应直方图均衡法增强视网膜图像;然后,采用不同尺度的二维Gabor小波对眼底图像进行变换;最后,采用广义线性模型(generalized linear model,GLM)分类器对眼底图像进行血管分割。
[Abstract]:Fundus vascular network is a deeper microvessel that can be observed directly by non-invasive method. Any systemic and hematologic lesions can lead to changes in the characteristics or morphology of the fundus microvessels. Ocular fundus image vascular segmentation is a key step in retinal image processing and analysis. It is of great significance for early prevention and diagnosis of systemic and hematologic diseases. The characteristics of retinal images are complex, and the automatic segmentation of fundus vessels is easy to be affected by the external conditions and the pathological changes themselves. Moreover, in the fundus images, the contrast between the tiny vessels and the background is low, which increases the difficulty of the segmentation of the fundus vessels. Therefore, improving the segmentation accuracy of ocular fundus vessels is an important research topic. This paper introduces the background and significance of the study of fundus vessels and the structural characteristics of the eyeball and retina, expounds the present situation of the research on the image processing of the fundus at home and abroad, and analyzes the properties and difficulties of the image of the blood vessel. In this paper, the image denoising of fundus and the segmentation of blood vessel in fundus image are studied. In this paper, the color fundus image in DRIVE standard image database and STARE standard image database are used for simulation experiment. The main works are as follows: (1) an improved bilateral filtering method for denoising fundus images with non-local mean filtering is proposed. In this paper, the denoising of fundus image is studied, and the two denoising methods of bilateral filtering and non-local mean filtering are deeply analyzed, and the advantages and disadvantages of the two methods are summarized. On the basis of this, an improved two-sided filtering method for denoising the fundus images with non-local mean filtering is improved. At the same time, the integral image operation method and the raised cosine function approximate gray similarity function are used to realize the fast operation of NLMF and BF, respectively. The simulation results show that the improved de-noising method has a good effect on the image denoising of the fundus, and the computation time is less. Based on the analysis and study of the fundus blood vessel segmentation method, a color fundus blood vessel segmentation method based on the level set function is studied in this paper. At first, adaptive histogram and two-dimensional Gabor transform are used to preprocess the fundus image, and then the level set theory of variable region fitting energy is used to simulate the blood vessel segmentation of the fundus image. According to the characteristics of fundus vessels, a color fundus image segmentation algorithm based on generalized linear model is improved. The algorithm firstly uses adaptive histogram equalization method to enhance retinal image, then uses two-dimensional Gabor wavelet to transform the fundus image. Finally, the generalized linear model (generalized linear model-GLM) classifier is used to segment the blood vessel of the fundus image.
【学位授予单位】：广西师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R770.4;TP391.41

【参考文献】