医学图像分割与重建网格模型简化的研究

【作者】 胡永杰；

【导师】 解梅；

【作者基本信息】 电子科技大学 ， 信号与信息处理， 2008， 硕士

【摘要】 医学影像处理与分析是目前的一个研究热点问题,是一个多学科交叉的研究领域,是计算机图形学和图像处理在生物医学工程中的重要应用。它涉及数字图像处理、计算机图形学以及医学领域的相关知识。医学影像处理与分析的主要研究内容包括医学图像的预处理;组织或器官的分割与提取;复杂表面多相组织成份三维几何模型的构建;重建模型的表面网格简化;模型的剖切与手术开窗操作等。本文研究的主要内容为医学图像的预处理、组织或器官的分割和三维重建模型的网格简化。分析和总结了国内外学者在这三个方面的研究,本文研究和探讨了几种新的方法,改善现有常用方法存在的不足。本文的创新点和独立工作主要体现在以下几个方面:1.研究和总结了医学影像处理与分析中的医学图像的预处理技术、医学图像的分割技术和虚拟可视化中的三维模型简化技术的现状和存在的问题;2.改进了各向异性扩散的滤波方法,解决了传统滤波方法在去除噪声的同时,也去除了高频边缘信息的问题,同时改进了各向异性扩散滤波滤波速度慢的不足。实验结果证明了该方法的优点。3.研究了基于决策树的头颅MRI图像的分割方法。该方法结合了模式识别、传统图像分割技术和头颅组织空间分布的规律,能自适应的分割颅内组织,实验结果表明该方法是可行而且有效的。4.研究了医学图像序列交互式的分割方法。该方法采用分数阶梯度来改进传统的live wire交互式分割方法,然后运用轮廓插值技术对图像序列进行自动快速的分割。实验结果充分体现了该方法的优点。5.改进了经典的顶点抽取模型简化方法,重新定义了顶点重要性的度量,提高模型简化的质量,结合自适应八叉树的优点,实现三维模型快速简化。实验结果证实了该方法的优点。更多还原

【Abstract】 Medical image processing and analysis is the one of popular research projects, which is the project including much different knowledge and is the application of computer graphics and image process in the biomedical engineering. It covers many subjects such as computer graphics, image process and medical knowledge, and it is mainly constituted by the pre-processing of medical image, the segmentation of the apparatus and the tissue, visualization of 3D medical models, simplification of complex models, dividing cubes of models and so on.The primly contents of this thesis are pre-processing of medical image, image segmentation and the simplification of complex models. After studying and summing up many dissertations which obtained from the domestic and overseas scholars, we propose several new arithmetic, to improve the shortages of the classical’s.The key work and innovations of this thesis mainly include:1. Research of the actuality of the pre-processing of medical image, image process and the simplification of complex models.2. Proposed a improvement of anisotropy filter in pre-processing of medical image, which not only wipes off the noise, but also keeps the high-frequency edge information, and improves the time-consuming shortage. Experiments show its advantage.3. Presented a new approach of skull’s MRI segmentation based on decision making tree. Combining the pattern recognize and the distributing rule of tissue in the skull, it can auto-adapting segment the objects. Experiments show it is feasible and effective.4. Presented a improvement approach in medical image mutual segmentation. It improves the live wire arithmetic through using Fractional Differential to replace the integral grads, then segments the medical image series automatically by contour interpolation. Experiments show its advantages.5. Proposed a improvement of vertex deletion in medical model simplification. It defines a new measurement of vertex essentiality, to advance the quality of simplified models; and uses the advantage of auto-adapt octree to quicken up the speed of the model simplification. Experiments show its advantages.更多还原