【作者】 康晓东；

【导师】 何丕廉；

【作者基本信息】 天津大学 ， 计算机应用技术， 2008， 博士

【摘要】 合理地分割影像图像是医学影像图像处理的重要基础步骤之一。鉴于医学影像的特殊性,常采用蚁群算法和形变模型法等基于人工生命的图像分割方法。本文对蚂蚁信息激素释放、路径转移重新定义,并将图像空间的模糊连接关系引入蚁群的觅食过程中,进而转换为蚁群搜寻的目标食物,改进了基于蚁群算法的医学图像分割方法。本文将Live Wire引入分水岭图像分割算法,对相邻区域以其像素数、灰度均值和灰度方差定义距离,并据其建立新连通图,以对图像过度分割而产生的一些过小区域合并;通过增加灰度模型,并以灰度模型中已分割图像中包含的物体和背景的局部区域统计信息,以及距离图等信息,引导活动轮廓在相邻的待分割图像中快速准确地收敛到物体的实边界;在模型跨边缘时,利用已分割图像中模型内部区域的统计特征,用区域生长法获取内点并重新参数化模型,以保证模型形变到正确的边缘。从而改进了基于形变模型的医学图像分割。本文面向PACS应用及不同类型物理存储器的技术特点,多层配置图像存储,以虚拟设备管理器屏蔽存储介质物理上的差异;针对医学图像传输时间分布特征,采用基于最小频率-时间间隔比的图像信息调度策略,并给出具体算法。本文分别对改进的医学影像图像分割结果进行了对比测试。更多还原

【Abstract】 One of the most important basic processes of dealing with medical image is to segment it reasonably. Some artificial-life-based image segmentation methods, such as Ant Colony Algorithm and deformable model, are usually applied to medical image segmentation.This thesis re-describes the ant pheromone releasing and its transferring route. In this thesis, a fuzzy connectedness of image is introduced to the process of ant’s seeking for food and, farther, transferred to the target food of art’s seeking for food. Thus, the method for the segmentation of medical image based on ant colony algorithm is improved in this thesis.By introducing the Live Wire to watershed medical-image segmentation, the author modifies the traditional live wire algorithm by redefining the distance of the neighborhood, using its pixel, gray-scale mean and gray-scale variance. Then a new connected graph is constructed by the computer, which combines the over- small-regions that is caused by over-segmenting to medical image. By appended gray-scale model and using the information of partial statistic and information of the distance to the image, which are concerned with the objectives and background that are included in the lines of segmented tomography image, the snake that lines in the neighbored and segmentation-waited tomography image can be led to converge to the factual boundary rapidly and accurately. The author also presents a new method for solving the problem of model’s crossing edge based on T-snake model. While the model is in crossing edge, we use statistical features of model’s interior region in the segmented tomography image, get the interior points by using region growing segmentation and reconstruct the parametric model In order to ensure the model to be deformed to the right edge.On the basis of applying PACS and considering the technical characteristics of different types of physical archiving medium, this thesis uses multilayer archiving configure and using suppositional equipment manager shields the physical differences between archiving mediums. Aiming at the distributing characteristics of medical-image transferring, the author adopts several image information attemperring strategies and concert algorithm, which is based on the spacing ratio between minimum frequency and time.Besides, the contrast testing is made separately for the result of improved medical-image segmentation in this thesis.更多还原