医学可视人真彩图像分割技术研究

【作者】 张雯；

【导师】 郝重阳；

【作者基本信息】 西北工业大学 ， 信号与信息处理， 2005， 硕士

【摘要】 医学图像分割是把医学图像分成各具特性的区域并且提取感兴趣目标的过程。它是医学图像处理和分析中的关键技术,也是目前制约其发展和应用的瓶颈。 由美国发起的可视人计划是一项人类借助计算机技术进一步认识自身,同时在医学、仿生学等多个领域具有广阔应用前景的重大课题。目前人体可视化研究进入到对人体及组织器官的重建和理解阶段,其中对真彩图像的组织器官的成功分割是该项目能否顺利进行的核心环节,同时也是目前最具挑战性的问题。 本论文应用可视人体真彩数据集中的胸部切片图像真实数据,着重研究了医学图像分割的基于区域的分割方法,基于边界的分割方法、区域和边界结合的混合分割方法,对其中经典的算法给出了试验结果。此外还研究了医学图像分割技术的评价标准。 在以上研究和分析的基础上,结合待分割目标-肺部器官的颜色特征,取得了关于医学图像分割的具有新意的成果: (1) 提出了一种基于自适应阈值选取和彩色空间特征的区域增长方法。理论分析和实验结果表明,比传统的区域增长方法,分割结果从主观视觉效果上有显著改善。 (2) 对于散见报导的“区域和边界相结合的混合分割方法”,从理论、算法、和实现方法等方面进行了系统的分析和归纳,得出了有供参考的结论性意见。 (3) 提出了一种改进的基于模糊连接度和Voronoi图的混合分割算法。将原来仅限于灰度图像的分割方法引申到彩色图像分割,并达到较好的效果。更多还原

【Abstract】 Medical image segmentation is a technology of dividing the image into segments of various characteristics and abstracting the target interested in. The technology is the focus and key point in medical image processing and analysis, and also the bottleneck, which restricts the development and application of medical image processing.Virtual Human Project, which human further recognize themselves by means of computer and launched by USA, is a significant research topic with wide application value in many fields such as medicine, bionics. At present, the research of human visualization has come into the moment of reconstruction and understanding of body and tissue organ, where the successful segmentation of organs in true color image is the core step, which decides whether the project can progress all right, and furthermore, is the most challenging task in VHP.This paper, using the true data of human thorax slice image from Virtual Human true color data set, studied the following methods of medical image segmentation: methods based on region, methods based on boundary, and methods based on hybrid segmentation—combining region and boundary. Experimental results were given for some classical algorithms. Furthermore, this paper analyzed in-depth the existing medical image segmentation evaluation methods and criteria.Based on above primary work, this paper obtained a few creative achievements about medical image segmentation:(1)Presented a new region-growing algorithm based on adaptive threshold selection and color space feature. Theory analysis and experiment results show that, comparing with traditional region growing algorithm, the new method can improve remarkably the segmentation results from subjective visual effect.(2)Summarized systemically the current hybrid segmentation methods combining region and boundary, from the aspects of theory, algorithm and implementation.(3)Presented an improved hybrid segmentation algorithm based on fuzzy connectivity and Voronoi diagram. Extended the hybrid segmentation method from the old one that can only be used for gray image to the new one that can be used for color image. The experimental results show that this improved method is promising.更多还原