SAR图像多尺度配准的小波域等周割方法  

Multi-Scale Synthetic Aperture Radar Image Registration by Isoperimetric Cut in Wavelet Domain

作　　者：李婧[1] 田铮[1] 

机构地区：[1]西北工业大学理学院,陕西西安710072

出　　处：《计算机仿真》2011年第1期246-249,330,共5页Computer Simulation

基　　金：国家自然科学基金(60375003);航空科学基金(03I53059);西北工业大学科技创新基金项目(2007KJ01033)

摘　　要：研究在同一场景中不同时间和角度的图像配准问题,针对保证配准的精确性,给出了基于合成孔径雷达(SAR)图像小波域等周割多尺度分割的SAR图像多尺度配准方法。为了抑制噪声和配准精度,提出了SAR图像小波域等周割模型并给出模型的多尺度分割算法,以得到较为精确的轮廓特征,结合轮廓修正过程对每个小波分解层进行多尺度配准;并且,配准过程决定了分割中尺度因子的变化程度。小波分解的应用可有效降低搜索空间,在一定程度上加快配准速度,且使用修正后的轮廓特征信息,提高了配准精度。实验结果表明了方法的有效性,为图像配准技术提供了参考。The paper presents a method of multi-scale synthetic aperture radar （SAR） image registration based on multi-scale SAR image segmentation in wavelet domain. The model of multi-scale SAR image segmentation in wavelet domain was proposed and the algorithm of multi-scale segmentation was presented to obtain accurate contour feature, combining with contour refining process, the algorithm of multi-scale registration was carried out in scales of wavelet decomposition. Moreover, the registration process decides the ehange degree of scale faetor in segmentation process. In the proposed method, the use of wavelet decomposition can reduce the search space effectively to a certain extent and speed up the registration consequently, and the use of information of refined contour feature improves regis- tration accuracy. The results of experiments demonstrate the validity and potential of the method.

关 键 词：合成孔径雷达 图像配准 多尺度 小波域 等周割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]