基于小波域MRF模型的SAR图像分割

针对存在严重斑点噪声的合成孔径雷达(SAR)图像的分割问题,提出了一种基于小波域马尔科夫随机场(MRF)模型的算法.该算法综合利用了隐含马尔科夫树的相关邻域信息和图像的多分辨率描述的信息,将期望最大化用于先验概率分布参数的估计,采用最大后验准则来进行图像的分割.通过对SAR图像的分割实验表明,该算法可有效去除斑点噪声的影响,并能在准确分割目标的同时保留目标的细节信息.

SAR image segmentation using MRF model in wavelet domain

To solve the problem of synthetic apertur e radar (SAR) image segmentation with multiplicative nature of the speckle. Therefore an SAR image segmentation algorithm using Markov random field (MRF) model in wavelet domain is proposed, in which the information of adjacent pixels of hidden Markov tree (HMT) and the information of multi-scale described images are combined. The expectation-maximization (EM) method is used for estimation of pre-probability; the maximum a posteriori (MAP) is used for segmentation of images. Si mulation shows that this algorithm can avoid the influence of speckle and achieve an accurate segmentation with detail information of the targets.