一种基于偏微分方程的SAR图像去噪方法

A New Speckle Noise Removal Method Based on PDE's

传统的相干斑噪声抑制算法在多次迭代后通常会导致图像边缘的模糊,这一直是SAR图像去噪处理的难点和热点所在。该文分析了应用于图像处理的各向异性扩散方程(PDE's),在其基础上由最小化问题出发,引入棱边指示子对图像的边缘加以限制,得到新的去噪模型并降之应用于SAR图像的相干斑噪声去除。与传统的基于局部统计量和各向异性滤波器相比,新的算法在棱边保持和噪声去除能力均有提高。

Traditional speckle removal methods generally result in the lost of image features. This paper presents a new approach to edge-preserving smoothing of digital SAR images based on the PDE＇s. A brief analysis of the principle of anisotropic nonlinear diffusion equation for image restoration is introduced, and then an edge indicator is showed in the final update equation to constrain the image edge, staying as close as possible to the input image and to restore discontinuities, which improve the ability of speckle noise removal. Comparison and experimental result show the new proposed algorithm has high performance.