一种基于新的相似性测度的自动图像配准算法  

Automatic image registration algorithm based on a novel similarity metric

作　　者：梅跃松[1] 杨树兴[1] 莫波[1] 

机构地区：[1]北京理工大学机电工程学院,北京100081

出　　处：《仪器仪表学报》2007年第S1期336-339,共4页Chinese Journal of Scientific Instrument

基　　金：国防基础科研基金(K1305060614)资助项目

摘　　要：图像配准是图像处理的一个重要技术,被广泛地用于遥感图像、医学影像、视频监控等诸多领域中。本文提出了一种基于新的相似性测度的配准算法,首先分析2幅图像的联合直方图点集的分布情况,基于此定义了直方图点集的散度公式,并将其作为相似性测量。为加速参数的搜索过程,配准是在小波域内进行的,并使用Powell算法来优化搜索策略。实验证明,相对于基于互信息的图像配准算法,本算法参数优化方法选择可以更灵活,时间消耗更少,噪声鲁棒性更优。Image registration is a kind of important technology of image processing,and is used in the field of remote sensing,medical image,video surveillance,and so on.An image registration algorithm based on a novel similarity measure was proposed,and the point set distribution of joint histogram was analyzed.Its divergence formula was defined and used as the similarity metric.To speed up searching the registration parameters,all were done in the wavelet field and Powell method was used to optimize this process.Exp...

关 键 词：图像配准 直方图散度 互信息 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]