基于图像局部几何结构的SAR图像降噪与增强*

研究了基于图像局部几何结构对SAR图像进行各向异性扩散滤波降噪。首先，回顾各向异性扩散滤波的PM模型、Weickert模型和Tschumperle的迹模型，分析指出迹模型能够依据图像局部几何结构进行定向扩散滤波且扩散程度由扩散率函数决定，扩散过程可控，意义直观；继而，根据扩散系数的构建原则，构建了新的兼容图像增强、扩散幅度可调整的扩散率函数，并用于SAR图像降噪。实验结果表明，运用此函数不仅有效抑制了相干斑噪声，还保持并增强了边缘细节，取得了理想的效果。

De-noising and enhancement for SAR image based on local geometric structure of image

This paper investigated anisotropic diffusion for SAR image based on local geometric structure of image. Firstly, it reviewed three anisotropic diffusion models: PM model, Weickert model and Tschumperle trace-based model and pointed out that the trace-baced model which could perform directional diffusion according to local geometric structure and whose diffusion ability depends on diffusivity functions was more controllable and meaningful. Then, following the principle of diffusion coefficients, it proposed new diffusivity functions combining image enhancement and diffusion control for SAR speckle reduction. The experimental results show that applying new diffusivity functions not only reduces SAR speckle efficiently but also enhances edges and details and achieves ideal effect.