摘要
针对SIFT应用于SAR图像同名点自动提取的问题,该文提出了一种新的基于各向异性尺度空间的同名点提取方法。该方法首先基于双边滤波器建立图像的各向异性尺度空间,在滤除斑点噪声的同时保留了图像细节;然后利用SIFT在大尺度上检测并描述特征,弱化了斑点噪声对匹配的影响;最后采用双向匹配策略建立同名点,提高了正确匹配的概率。该方法在保持同名点精度的同时增加了同名点的数量。通过不同时相、不同极化、不同波段以及不同视角下的SAR图像同名点提取实验,验证了该方法的优越性和适应性。
A novel method based on anisotropic scale space is proposed to find matches in SAR images,which overcomes the disadvantages of SIFT.First,the anisotropic scale space of the image is constructed using bilateral filters,which removes the speckle while preserving the detailed.Then features are detected and described using SIFT at coarser scales,which reduces the impact of speckles.Finally,matches are found by dual matching strategy,which increases the probability of correct matching.This method increases the number of correct matches while maintaining the accuracy.Through various experiments including slant range images acquired from different times,polarizations,wavelengths and viewpoints,it demonstrates the improvement over SIFT in term of the amount and accuracy of the matches.
出处
《电子与信息学报》
EI
CSCD
北大核心
2012年第2期287-293,共7页
Journal of Electronics & Information Technology
基金
国家863计划项目(2007AA120302)资助课题