一种基于外形区域的图像配准方法与实现  

Contour-Based Registration Algorithm and Its Implementation

作　　者：衣文文[1] 杨彬彬[1] 胡彦磊[1] 常发亮[1] 

机构地区：[1]山东大学控制科学与工程学院,山东济南250061

出　　处：《计算机技术与发展》2008年第4期1-4,8,共5页Computer Technology and Development

基　　金：山东省重点自然基金资助项目(Z2005G03)

摘　　要：医学图像配准是医学图像处理与分析的重要组成部分。但由于医学图像的复杂性,在配准精度和配准速度两方面均优的算法很少。提出了一种基于分割所得二值图像的配准算法,并提出了一种适合二值图像配准的相似化测度,最后详细讨论了此算法的实现问题。与现今最流行的基于互信息的配准方法相比,文中的算法更易于实现,且具有良好的特性:适用于单模和多模图像配准;亚像素级精度;快速的配准速度。The medical image registration is an important part of medical image processing and analysis. However, little registration algorithm has both precision and high speed due to the complexity. In this paper, a novel registration algorithm was presented, which analyzed a new similarity measure appropriate to align two-valued images, based on a segmented two-valued Image. The experimental results show that the method is easier to implement and faster in computation than other methods.

关 键 词：Intrasubject配准 区域增长 分割 相似性测度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]