基于模糊聚类的脑部MRI医学图像分割算法的研究与实现
发布时间：2018-09-04 06:42
【摘要】：现代医学成像技术中的磁共振成像(Magnetic Resonance Imaging,MRI)技术因为它的特点使得其在临床医学方面的应用需求不断地增加。并且由于计算机科学技术的发展使医学图像更加广泛的应用于医学生物研究和临床医学方面。然而,因为脑MRI图像中存在有部容积效应、噪声、灰度不均匀性和对比度低等缺陷,在实际应用中脑MRI图像的精确分割是十分困难的。据此,本文基于基础模糊C均值聚类算法,针对MRI医学图像中存在的一些问题,对算法提出一定的改进,提出了一种基于各向异性滤波的模糊C均值聚类算法、进行了灰度不均匀场的移除工作的研究并且应用到实际脑MRI图像的分割工作中去。本文主要做了以下几方面的工作:(1)结合各向异性滤波去噪的方法提出了提出了一种基于各向异性滤波的模糊C均值聚类算法(PMFCM)。首先在本论文中研究分析基础模糊聚类算法的构成及其主要缺陷,随后研究了现有算法的改进思路,通过引入各向异性滤波方法对图像进行滤波得出的各中心点来代替空间邻域项的构造,不仅使得算法不用每次迭代时都计算其邻域信息使得迭代时速度加快,并且由于采用各向异性滤波处理后的像素作为约束项的计算信息。本文算法可以有效的抵抗待分割图像的噪声,使得分割效果更加可靠,并且利用快速模糊C均值聚类算法思想,使得本文算法的运算效率得到提高。(2)研究了含灰度不均匀场的脑MRI图像的分割策略,并且将其引入到本文所提出的基于各向异性滤波的模糊C均值聚类算法中。使得本文算法可以有效的估计出图像的灰度不均匀场的信息,利用在分割进行的同时估计灰度不均匀场的方法,得到更加清晰的移除灰度不均匀场的分割后图像。(3)研究了基于本文算法的聚类数目的初始化问题。在本文算法对图像进行各向异性滤波的前提下,利用二次差分的方法得到初始聚类中心和聚类数目,进一步加快了本文算法的分割速度。(4)根据本文算法设计并实现了医学图像处理系统图像分割子系统。
[Abstract]:Magnetic resonance imaging (Magnetic Resonance Imaging,MRI) technology in modern medical imaging technology has been increasing the demand for clinical applications because of its characteristics. Because of the development of computer science and technology, medical images are widely used in medical biological research and clinical medicine. However, due to the defects of partial volume effect, noise, gray inhomogeneity and low contrast in brain MRI images, it is very difficult to segment the brain MRI image accurately in practice. Therefore, based on the basic fuzzy C-means clustering algorithm, a fuzzy C-means clustering algorithm based on anisotropic filtering is proposed to solve some problems in MRI medical images. The work of removing the gray inhomogeneous field is studied and applied to the segmentation of the actual brain MRI image. The main work of this paper is as follows: (1) A fuzzy C-means clustering algorithm (PMFCM). Based on anisotropic filtering is proposed. In this paper, the composition of the basic fuzzy clustering algorithm and its main defects are studied, and then the improved ideas of the existing algorithms are studied. By introducing the anisotropic filtering method to replace the construction of spatial neighborhood terms, the algorithm not only computes its neighborhood information at every iteration, but also accelerates the iteration time. Due to the use of anisotropic filtering pixels as constraint information. This algorithm can effectively resist the noise of the image to be segmented, make the segmentation more reliable, and use the fast fuzzy C-means clustering algorithm. The computational efficiency of this algorithm is improved. (2) the segmentation strategy of brain MRI image with gray inhomogeneous field is studied and introduced into the proposed fuzzy C-means clustering algorithm based on anisotropic filtering. So that the algorithm can effectively estimate the information of the image gray inhomogeneous field, and use the method to estimate the gray level non-uniform field while the segmentation is going on. The segmented image with a clearer removal of the inhomogeneous gray field is obtained. (3) the initialization problem of the clustering number based on the proposed algorithm is studied. On the premise of anisotropic filtering of image in this paper, the initial clustering center and the number of clustering are obtained by using the method of quadratic difference. The segmentation speed of this algorithm is further accelerated. (4) the image segmentation subsystem of medical image processing system is designed and implemented according to this algorithm.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：R445.2;TP391.41
本文编号：2221316
[Abstract]:Magnetic resonance imaging (Magnetic Resonance Imaging,MRI) technology in modern medical imaging technology has been increasing the demand for clinical applications because of its characteristics. Because of the development of computer science and technology, medical images are widely used in medical biological research and clinical medicine. However, due to the defects of partial volume effect, noise, gray inhomogeneity and low contrast in brain MRI images, it is very difficult to segment the brain MRI image accurately in practice. Therefore, based on the basic fuzzy C-means clustering algorithm, a fuzzy C-means clustering algorithm based on anisotropic filtering is proposed to solve some problems in MRI medical images. The work of removing the gray inhomogeneous field is studied and applied to the segmentation of the actual brain MRI image. The main work of this paper is as follows: (1) A fuzzy C-means clustering algorithm (PMFCM). Based on anisotropic filtering is proposed. In this paper, the composition of the basic fuzzy clustering algorithm and its main defects are studied, and then the improved ideas of the existing algorithms are studied. By introducing the anisotropic filtering method to replace the construction of spatial neighborhood terms, the algorithm not only computes its neighborhood information at every iteration, but also accelerates the iteration time. Due to the use of anisotropic filtering pixels as constraint information. This algorithm can effectively resist the noise of the image to be segmented, make the segmentation more reliable, and use the fast fuzzy C-means clustering algorithm. The computational efficiency of this algorithm is improved. (2) the segmentation strategy of brain MRI image with gray inhomogeneous field is studied and introduced into the proposed fuzzy C-means clustering algorithm based on anisotropic filtering. So that the algorithm can effectively estimate the information of the image gray inhomogeneous field, and use the method to estimate the gray level non-uniform field while the segmentation is going on. The segmented image with a clearer removal of the inhomogeneous gray field is obtained. (3) the initialization problem of the clustering number based on the proposed algorithm is studied. On the premise of anisotropic filtering of image in this paper, the initial clustering center and the number of clustering are obtained by using the method of quadratic difference. The segmentation speed of this algorithm is further accelerated. (4) the image segmentation subsystem of medical image processing system is designed and implemented according to this algorithm.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：R445.2;TP391.41
【参考文献】
相关期刊论文 前1条
1 王兴;冯子亮;;基于自适应初始值的FCM聚类图像分割[J];计算机技术与发展;2010年03期
，本文编号：2221316
本文链接：https://www.wllwen.com/yixuelunwen/fangshe/2221316.html
