图像非刚性配准技术及并行策略研究  

Research on the Serial and Parallel Strategies for Non-Rigid Image Registration

作　　者：杜云飞[1] 王攀峰[1] 周海芳[1] 贾佳[1] 杨学军[1] 

机构地区：[1]并行与分布处理国家重点实验室,湖南长沙410073

出　　处：《计算机工程与科学》2007年第11期34-36,39,共4页Computer Engineering & Science

基　　金：国家自然科学基金资助项目(60621003;60603081)

摘　　要：图像非刚性配准是近几年配准方法的一个研究热点,由于其计算复杂度很高,它的并行化问题成为近几年来的研究热点。本文首先给出了图像非刚性配准的数学模型,并给出了形式化定义;其次分类总结了近年来图像非刚性配准串行算法研究的新进展,从而在此基础上重点讨论了相应的并行化策略。本文还详细分析了设计并行图像非刚性配准算法需要考虑的几个问题;并比较评价了现有并行算法的性能特点,得出了一些结论;最后提出了有待进一步研究的问题。Non-rigid image registration is a hot topic of image registration in mathematical morphology, and its parallelism is an important problem. This paper starts with a mathematical model of non-rigid image registration in topography, followed by its formal definition. Then some new serial non-rigid image registration algorithms proposed in recent years are classified and analyzed,based on which the related parallel strategies are emphasized in detail. Then we analyze some issues, which should be considered when designing parallel non-rigid image registration algorithms, and give an evaluation of the existing parallel algorithms, and draw some conclusions. Finally we point out the problems and challenges of future research.

关 键 词：非刚性配准 区域匹配 并行算法 加速比 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]