基于小波不变矩的医学图像配准技术研究  

New Technique Research of Medical Image Registration Based on Invariant Moments

作　　者：董卫军[1] 樊养余[2] 刘晓宁[1] 周明全[1] 

机构地区：[1]西北大学信息科学与技术学院,西安710069 [2]西北工业大学电子信息学院,西安710072

出　　处：《计算机科学》2008年第7期234-236,共3页Computer Science

基　　金：国家自然科学基金(60573179);陕西省科技计划工业科技攻关项目(2004k05-G24)的资助

摘　　要：图像配准是对取自不同时间、不同传感器或不同视角的同一场景的两幅图像或者多幅图像匹配的过程,是图像融合、目标探测识别和计算机视觉等技术的重要基础,主要用于消除来自不同传感器的图像中目标的位置差异。提出了一种新图像匹配算法,新算法以提升格式小波变换为基础,利用图像多级小波分解后近似分量的轮廓相似性进行图像的快速配准。实验结果表明,新方法高效精确。Image registration is the matching processing which matchs two or more images from the same scene derived from different tittle, different sensors. It is the key basement of image fusion, target detection and recognition system, and computer vision. It＇s mainly used to eliminate the variation of position of targets from different sensor images. Given a new image registration algorithm which is based on lifting scheme wavelet transform. Decomposed wavelet components of the original images are used to impose the speed of image registration. Experiment results show that the algorithm is effective and accuracy,which is adapted not only to the registration of the mono-modal medical image but also to the multi-modal medical image.

关 键 词：小波变换 图像匹配 提升格式 不变矩 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]