基于可调滤波改进互信息的遥感图像配准  

Remote sensing image registration based on mutual information improved by steerable filters

作　　者：史振广[1] 田铮[1,2] 

机构地区：[1]西北工业大学理学院,陕西西安710129 [2]中国科学院遥感应用研究所,北京100101

出　　处：《纺织高校基础科学学报》2009年第2期250-253,267,共5页Basic Sciences Journal of Textile Universities

基　　金：国家自然科学基金(60375003);航空基础科学基金(03I53059)

摘　　要：为了减少基于互信息的图像配准方法中插值效应对配准精度的影响,提出一种用可调滤波改进互信息的配准方法,通过用可调滤波提取图像的全部方位结构信息特征进行互信息的估计.理论分析和实验结果表明该方法可以明显减少插值效应的影响,提高配准的精度,对遥感图像的配准结果也令人满意.In order to reduce the interpolation artifacts which influence the registration accuracy in mutual information based image registration method, a new registration method based on mutual information improved by steerable filters was proposed, in which the steerable filters were utilized to pick up the whole orientation information feature of images to estimate the mutual information. The theoretical analysis and experimental results show that the method is able to reduce the local extreme of mutual information function and improve the registration accuracy greatly, and the remote sensing image registration results are satisfied.

关 键 词：图像配准 互信息 插值效应 可调滤波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]