一种参数自适应的SAR图像去噪方法

Parameter adaptive SAR image denoising method

传统的合成孔径雷达(SAR)图像非局部均值去噪算法中,对于噪声块的相似性度量都简化为噪声点之间的相似性级联.在加性噪声模型下取得了很好的去噪效果.笔者将这种思想推广到了SAR图像的乘性噪声模型下,并且将其在最大似然权重估计的框架下对基于概率斑点(PPB)进行了改进.由于在PPB算法中参数的设置复杂,且不能自适应地获得最优效果,文中提出了基于粒子群优化的参数自适应的SAR图像去噪非局部算法.最后,在真实的SAR数据上,对该算法进行了实验测试,并与经典的PPB算法进行了实验对比.实验验证了该算法能更好地抑制噪声并且同时保持细节信息.

In the traditional SAR image nonlocal means denoising algorithms,the patch similarity is measured by the accumulation of the pixel similarities,and a good denoising performance can be obtained for the additive noise model.This paper extends this idea to the multiplicative noise model for the SAR image,and improves the PPB（Probabilistic Patch-Based）algorithm under the weighted maximum likelihood estimation framework.Since the parameters setting in the PPB algorithm is complicated and it cannot adaptively get the best performance,this paper proposes a particle swarm optimization based parameter adaptive nonlocal means algorithm for SAR image denoising.Finally,experiments compared with the canonical PPB method on the real SAR image are carried out.Experiments demonstrate that the proposed method has a good performance in speckle reduction and details preservation.