同时，我们探讨了基于图像表面轮廓的配准方法，阐述了基于表面轮廓的配准算法 的原理和要点。采用 Canny 算子进行表面轮廓提取，基于四元数和奇异值分解的方法进 行空间对应点的配准以及最邻近迭代搜索(ICP)。为了提高配准速度，我们采用了多步 长迭代的方法，实验证明该方法是一种比较可靠的方法。 关键字： 三维医学图像，图像配准，互信息，表面轮廓，优化，插值算法
ABSTRACT.......................................................................................................................................................... II
分类号： TP391 UDC： 004.9
学号： 020972
东南大学 硕士学位论文
三维医学图像配准方法研究
研究生姓名：江 军 导 师姓名：舒华忠 教授
於文雪 副教授
申请学位级别 工学硕士
学科专业名称 生物医学工程
论文提交日期 2005 年 月 日 论文答辩日期 2005 年 4 月 5 日
学位授予单位 东 南 大 学 学位授予日期 2005 年 月 日
1.1 引言 .............................................................................................................................................................. 1 1.2 医学图像配准.............................................................................................................................................. 2 1.3 本文内容组织.............................................................................................................................................. 3
The appearance of different kinds technology of imaging makes people easy to observe Dissection structure of human body. It is very helpful, but information of image that receives from the same human body is difference and complement to another. To make full use of a large amount of information, we must come to explain and utilize them as a whole, it is very important and useful in clinical diagnosis.
In the thesis, we first discuss registration by maximization of mutually information. It is fully automatic and images involved in registration don't need any pre-treat. We discuss principle of registration by maximization of mutual information at aspect of entropy and mutual information. We emulate key points involved in registration by maximization of mutually information in detail, such as transformation, interpolation methods and optimization procedure etc. We analyse the impact of PV on mutual information. In order to improve speed, we use many skills to raise speed, such as multi-resolution, reduction of grey levels and Hybrid Powell and Simulated Annealing optimization Algorithm etc, experiment shows that it is valuable to avoid local extremum.
研究生签名：
日 期：
东南大学学位论文使用授权声明
东南大学、中国科学技术信息研究所、国家图书馆有权保留本人所送交学位论文的 复印件和电子文档，可以采用影印、缩印或其他复制手段保存论文。本人电子文档的内 容和纸质论文的内容相一致。除在保密期内的保密论文外，允许论文被查阅和借阅，可 以公布（包括刊登）论文的全部或部分内容。论文的公布（包括刊登）授权东南大学研 究生院办理。
Then, we discuss the surface based registration, explain the principle of registration based on surface and enumerate the main points involved in surface based registration. We extract the contour by canny algorithm, discuss Quatemions and SVD which is used in the points-based registration, and Iterative Closest Points algorithm which is frequently used Iterative algorithm. In order to raise speed, we use multi-scale iterative algorithm.
BY Jiang Jun
Supervised by Prof. SHU Hua-zhong
Prof. YU Wen-xue
Department of Biomedical Engineering Southeast University March 2005Biblioteka 东南大学学位论文独创性声明
本人声明所呈交的学位论文是我个人在导师指导下进行的研究工作及取得的研究成 果。尽我所知，除了文中特别加以标注和致谢的地方外，论文中不包含其他人已经发表 或撰写过的研究成果，也不包含为获得东南大学或其它教育机构的学位或证书而使用过 的材料。与我一同工作的同志对本研究所做的任何贡献均已在论文中作了明确的说明并 表示了谢意。
Key Words: medical image, image registration, mutual information, surface, timization,
interpolation
东南大学硕士学位论文
目录
摘 要 .....................................................................................................................................................................I
答辩委员会主席 罗立民 教授 评 阅 人
2005 年 3 月 20 日
3D MEDICAL IMAGE REGISTRATION
A Dissertation Submitted to Southeast University
For the Academic Degree of Master of Engineering
The registration in medical image is an important technology in the medical image processing, it is a multi-disciplinary research fields that crosses information science, image technology of the computer and medical Science. The present medical image registration can be divided into methods based on the grey levels and methods based on characteristic of the image.This paper discusses registration by maximization of mutually information and surface based image registration, these two kinds methods are mostly used3D registration method in clinic.
东南大学硕士学位论文
Abstract
Title: 3D Medical Image Registration Graduate Student: Jiang Jun Thesis Supervisor: SHU Huazhong &YU Wenxue School: Southeast University