结合空时上下文信息的视频SAR图像相干斑滤波

VideoSAR image speckle filtering with space-time context information

作为合成孔径雷达(synthetic aperture radar,SAR)的一种新体制,视频SAR能够对场景连续成像,在目标检测、跟踪等应用情景有独特优势。然而,相干斑的存在影响了视频SAR图像质量及后续目标检测应用。针对这一问题,提出了一种结合空时上下文信息的视频SAR图像相干斑滤波方法。该方法主要由3步构成:首先构建能表征视频SAR图像距离-方位-时间三维动态散射信息的空时上下文协方差矩阵;其次根据协方差矩阵的相似检验选取一定邻域内目标像素的相似样本;最后根据相似样本进行平均实现相干斑滤波。基于两组实测视频SAR数据开展相干斑滤波对比实验,实验结果表明所提方法能在有效抑制相干斑的同时较好地保留目标细节信息,验证了所提方法的有效性和优势。

As a new system of synthetic aperture radar(SAR),VideoSAR can continuously image the scene,and has unique advantages in target detection,tracking and other application scenarios.However,the existence of speckle affects the quality of VideoSAR image and subsequent application of target detection.To solve this problem,a VideoSAR speckle filtering method combined with space-time context information is proposed.The method mainly contains three steps.Firstly,space-time context covariance matrices which can characterize the range-azimuth-time three-dimensional dynamic scattering information of VideoSAR image pixels are constructed.Secondly,the similarity samples of considered pixels are selected within a certain neighborhood according to the similarity test of context covariance matrix.Finally,speckle filtering is realized by averaging the similar samples.Based on two groups of measured VideoSAR data,speckle filtering comparison experiments are carried out to verify the effectiveness and advantages of the proposed method.It is proved the proposed method is effective in suppressing speckle while preserving further target’s detailed information.