一种基于相关邻域模型的SAR图像迭代滤波方法

SAR Image Iterative Filter Method Based on Correlated Neighbourhood Model

提出一种基于相关邻域模型的 SAR图像迭代滤波方法。该方法引入相关邻域模型描述图像的局部特性。首先建立一系列反映方向信息的模板 ,每一个模板代表一种邻域结构。在 SAR图像的统计性质基础上 ,对于每个邻域结构 ,利用 Bayes方法 ,根据图像的观测值 ,获得强度期望值的最大后验概率 (MAP)估计。运用模拟退火算法 ,根据每个结构的概率 ,随机选取一个邻域结构作为最适合结构 ,它所对应的估计值作为期望强度的估计。温度逐步趋向于 0 ,该算法可以获得估计的全局最优解。

A kind of SAR image iterative filtering methods based on correlated neibourhood model is proposed. This method describes a local image property by a correlated neighbourhood model. Firstly, a series of templates reflecting direction information are established and every template represents a kind of neighbour structure. Then based on the statistical property of SAR image, by using Bayes formulation the MAP estimate of the real intensity by observation image values for every structure is obtained. Simulated annealing algorithm is used to get proper structure from the candidate structure according to their probability, the estimate value corresponding this structure is the estimate of the underlying cross-section. The global optimum result can be obtained by simulated annealing algorithm when the temperature is reduced to zero.