Abstract: Coherent optical images often suffer from speckle noise.An anisotropic diffusion scheme is proposed to tackle the problem of image restoration in coherent optical image processing.The diffusion coefficient in the proposed algorithm theoretically satisfies the conditions for the design of diffusion coefficients presented by Charbonnier，et al.At the same time，this paper combines the modified anisotropic diffusion scheme with logarithmic transformation to restore the original appearance of coherent image.Experimental results show that compared with the SRAD，Lee，Frost，Non-local Means，BLS-GSM and BM3D method，the proposed anisotropic diffusion scheme can effectively remove speckle noise in coherent optical image and preserve the edge and texture features.

Key words: anisotropic diffusion, speckle, diffusion coefficient

摘要： 相干光学系统中图像噪声污染严重，提出一种基于各向异性扩散的图像复原模型，该模型的扩散系数从理论上满足Charbonnier等人提出的构造扩散系数准则，同时结合对数变换对相干光学图像进行复原去噪处理。实验结果表明，同SRAD、Lee、Frost、Non-local Means算法和基于小波的BLS-GSM、BM3D等方法相比，该算法可以更有效地进行边缘平滑，而且还能够较好地解决图像边缘和细节失真的问题，不论从PSNR，还是从视觉评价效果来看，该算法都具有一定的优越性。

关键词: 各向异性扩散, 斑点噪声, 扩散系数