摘要
针对传统合成孔径雷达图像变化检测方法检测精度低的问题,提出了一种在非下采样轮廓变换域结合显著图信息构造融合差异图的变化检测方法.首先用两幅输入图像分别构造均值比图、对数比图和邻域对数比图等3种差异图,并用对数比图获取显著图信息.然后对均值比图和邻域对数比图进行3级非下采样轮廓变换域分解,低频融合时对邻域对数比图的低频子带用显著图进行范围限定,以突出融合差异图的变化区域,高频融合时对两幅差异图的方向子带进行选择性的显著图限定去噪,再用局部能量最小原则进行融合,以抑制融合差异图的背景区域.最后经非下采样轮廓反变换得到融合差异图,对其进行k均值聚类,输出检测结果图.实验数据表明,文中方法能较好地保留边缘和细节信息,因而具备更高的图像变化检测精度.
Aiming at solving the problem of low detection accuracy of traditional synthetic aperture radar(SA R)image change detection methods this paper proposes a change detection method which combines salient information to construct the fused difference image in the nonsubsamplcd contourlct transform(NSCT)domain.First,three difference images including the mean ratio image,log ratio image and neighborhood log ratio image arc constructed with two input images and then the saliency image is extracted from the log ratio image Second,the mean ratio image and neighborhood log ratio image arc decomposed by the NSCT method.The low-pass sub-bands of the neighborhood log ratio image arc restricted by the saliency image to highlight the change region of the fused difference image.The directional sub-bands arc selectively denoised by the saliency image in different scales and then fused according to the principle of minimum local energy.Finally,the NSC1"inverse transform is used to obtain a fused difference image and the change detection map is generated by using the^-mcans clustering method.The experimental results show that this method can get a better edge and detailed information as well as a higher detection accuracy.
作者
慕彩红
吴生财
刘逸
彭鹏
刘若辰
MU Caihong;WU Shengcai;LIU Yi;PENG Peng;LIU Ruochen(Ministry of Education Key Lab.of Intelligent Perception and Image Understanding,Xiclian Univ.,Xi’an 710071,China;School of Electronic Engineering,Xidian Univ.,Xi’an 710071,China;School of Physics and Optoelectronic Engineering,Xidian Univ.,Xi’an 710071,China)
出处
《西安电子科技大学学报》
EI
CAS
CSCD
北大核心
2018年第2期19-25,共7页
Journal of Xidian University
基金
国家重点基础研究发展计划资助项目(2013CB329402)
国家自然科学基金资助项目(61672405,61373111,61573015,61473215,61371201)
中央高校基本科研业务费专项资金资助项目(JBG160229,JB170204)
国家留学基金资助项目(201706965003,201606965051)