【作者】 朱丙丽；

【导师】 陶宏才；

【作者基本信息】 西南交通大学 ， 计算机应用技术， 2010， 硕士

【摘要】 论文以医学领域的脑部CT (Computed Tomography)图像为研究对象,在医学影像存档与通讯系统PACS (Picture archiving and communication systems)的CT影像工作站环境下,研究了边缘分割算子中的LOG (Gauss-Laplacian Operator)算子、基于形态学分割的分水岭算法的应用和改进,获得了脑部CT图像中头骨的边缘分割图和头骨的区域分割图。论文根据医学图像的分割特点及方法,分析了脑部CT成像的原理、研究了脑部CT图像的特点、噪声、边缘特征、多阈值形态学特征。完成的主要工作如下：1.在高斯-拉普拉斯算子(LOG)的基础上,改进了根据高斯函数的平滑因子σ、脑部CT图像的边缘灰度阈值T、滤波器的大小m×n等三个参数控制的LOG算子,利用该改进的算子能根据需要灵活地得到二值化头骨边缘分割结果。2.以脑部CT图像的灰度多阈值、形态学特征为基础,应用去噪、距离变换、梯度重构、内外部标记条件等综合方法,改进了分水岭分割算法。利用该改进的分水岭分割算法能够得到脑部CT图像的头骨分割图,该改进算法在噪声抑制、区域信息获取和过分割控制等方面效果较好。3.利用Matlab语言实现了改进的LOG算子和改进的分水岭算法,并对它们的分割效果和运算效率作了分析。4.用改进LOG算子、改进分水岭算法对一组脑部CT切片图像进行了处理,综合分析了分割质量、运行时间和效率、复杂度等。论文中的改进算法,在理论上有较全面和深入的研究,在应用上获得了有效的分割结果,其可行性和有效性从理论与实践上得到了验证。更多还原

【Abstract】 Taking CT image of the brain as the researched subject, under the environment of CT workstation of PACS, the author of this thesis studied and improved the LOG and watershed algorithm in the stage of image segmentation, achieved the results of CT image of the brain on skull edge and region.According to the feature and approaches, the thesis studied the principle, characteristics, noise, edge feature, morphological feature of multi-threshold values in CT image of the brain’s skull segmentation. The main achivements are as follows:1. According to the feature of edge segmentating on skull, the LOG method is improved, and the binary result can be obtained flexiblely through the improved LOG agorithm, which is controlled by Gaussian function’s smoothing factorσ, edge gray threshold value T, and pepper-and-salt filter’s size m×n.2. According to the morphological feature of multi-threshold values, the watershed segmentation algorithm is improved by combining denoising, distance transform, gradient reconstruction with constraint conditions, and the segmentation image of brain skull which is obtained by the improved algorithm is good at noise suppression, region information accepting, and over segmentation controlling.3. The segmented images from the improved LOG and watershed algorithm realized with Matlab language can be achieved effectively, and the segmemting effect and algorithm efficiency are analyzed.4. Applied improved LOG and improved watershed algorithm in a group of brain CT slides, and then appraised segmented results, run time, efficiency and complexity and so on.The improved algorithms of this thesis are studied fully and deeply in theory, and can achieve effective segmentation results in application. So, the feasibility and validity of the improved algorithms are verified from theories and experimentations.更多还原