SAR图像相干斑噪声的各向异性扩散滤波算法

Algorithm for reducing speckle noise in SAR image by anisotropic diffusion

传统各向异性扩散滤波算法以梯度作为边缘检测与扩散系数生成的主要依据,当其用于抑制SAR图像中的乘性相干斑噪声时,存在对噪声敏感,边缘检测非恒虚警,扩散不均衡等问题,为此提出了一种新的各向异性扩散滤波算法。该算法利用平行窗通过比率运算生成边缘强度映射(ESM),然后利用ESM代替梯度作为边缘检测与扩散系数生成的主要参数,并将方向扩散与反向扩散结合对图像边缘进行去噪与锐化,从而有效克服了传统扩散滤波算法在处理相干斑噪声时存在的问题。实验表明:与传统扩散滤波算法相比,新算法在SAR图像去噪与边缘保护性能上均有提升,且锐化了图像边缘。

In traditional anisotropic diffusion filter algorithms, gradient is used as the main parameter to determine the edges of SAR image and produce nonlinear diffusion coefficients. When these algorithms are selected to reduce multi- plicative speckle noise in SAR image, there are several shortcomings, such as sensitivity to noise, inconstant false alarm rate in edge detection, and imbalance diffusion. So a new algorithm based on anisotropic diffusion is proposed. In the new algorithm, edge strength map（ESM） is firstly produced by ratio operations of parallel windows, and gradient is replaced by ESM to be as the main parameter to determine the edges of SAR image and produce nonlinear diffusion coefficients. Then direction diffusion and inverse diffusion are adopted to reduce speckle noise and sharpen edge on the edge of SAR image. Thus, when the new algorithm is used to removal speckle noise, the shortcomings of traditional diffusion filter algorithms can be overcome effectively. The experimental results show that compared with traditional diffusion filter algorithms, the proposed algorithm has better performance in terms of reducing speckle noise and preserving the edge of SAR images, and can realize edge sharpening of SAR image.