自交叉双边滤波的极化SAR数据相干斑抑制

相干斑抑制是极化SAR数据预处理的关键步骤。双边滤波是一种空域和值域滤波相结合的优秀边缘保持滤波算法。针对双边滤波在抑制极化SAR数据相干斑的不足,该文将改进的交叉双边滤波引入到极化SAR数据降噪领域,加入散射机制测度来扩展原权重核,根据SPAN图像的局域变差系数自动调整空间方差系数,利用参考图像来度量灰度值和散射机制相似性。实验结果表明:本文方法较经典滤波算法有更强的噪声平滑能力和更好的细节信息保持能力,在保持原数据极化散射信息方面也表现出良好的性能,这为基于极化SAR数据的后续应用提供了支持。

Speckle filtering for polarimetric SAR data based on self-cross bilateral filter

Polarimetric Synthetic Aperture Radar (SAR) data record an increased amount of scattering information of ground targets with fully polarimetric patterns. Thus, polarimetric SAR has become a key issue in development. Polarimetric SAR technologies have made significant progress in applications such as military and civil remote sensing. However, because of the limitation of coherent imaging, speckle in SAR and polarimetric SAR data seriously affects the relative information extraction. Therefore, speckle suppression is an important step in SAR and polarimetric SAR data processing. Subsequent applications with polarimetric SAR data could be supported. The Bilateral Filter (BF) that combines spatial closeness and gray similarity to suppress speckle is an excellent edge preservation filtering algorithm. This paper proposes an Improved Cross Bilateral Filter (ICBF) to resolve the deficiency of the BF in speckle suppression of polar metric SAR data. The reference image was imported to improve the gray similarity measurement accuracy of the original data, and a new strategy that sets two key parameters of bilateral filter based on the variance coefficient of filter window was presented to adapt to the change in the homogeneity degree of the filter window. The ICBF added scattering mechanism measurement to extend the original weight kernel, adjusted spatial closeness variance via the local coefficient of variation by using SPAN image, and measured the similarity of gray value and scattering mechanism by using a reference image. The experiment was conducted with AIRSAR polar metric data in the San Francisco region. The proposed filter algorithm was then compared with classic filter algorithms, such as Boxcar, IDAN, and Refined Lee, based on speckle suppression, detailed information preservation, scattering and polarization information preservation with the visual and index methods. The experimental results show that ICBF could improve the gray similarity measurement accuracy of the original data and that ICBF can suppress speckle and preserve detailed information, such as point targets and edges in the original data, better than the classic filter algorithms. By contrast, the measurement of scattering mechanism was combined in the kernel function of bilateral filter, the parameter of the space distance weight was obtained from SPAN image, and the weights of gray value and scattering mechanism were measured with the standard deviation of the corresponding data. Thus, the polarization information and scattering characteristics of the original data were preserved well. Compared with classic filtering algorithms, the ICBF can smooth noise and preserve detailed information better. Furthermore, polarization information and scattering characteristics of the original data are maintained to support subsequent application based on polarimetric SAR data.