SAR图像目标增强模型最优性能分析

研究用于SAR图像目标增强的lk范数正则化方法的最优性能.从有偏参数估计的统计性能极限角度,将达到渐进最小总方差的参数估计看作目标增强的最优效果.根据Bayes框架以及罚最大似然估计原理,研究了达到最优性能时正则项需满足的条件,该条件与图像大小、信号统计分布以及观测数据噪声方差有关,可为模型设计与迭代算法设计提供依据.并证明了仅当噪声方差较小时lk范数约束才能使得处理结果达到该最优估计性能.仿真实验验证了该结论的正确性.

Optimal estimation performance analysis for targets enhancement model of SAR image

The optimal performance of the regularization method based on lk norm for SAR image feature enhancement is investigated.From the view of statistical performance limit of biased estimation,the estimator whose total variance is least is regarded as the optimal efficiency.Based on the Bayesian framework and penalty maximum likelihood theory,the condition to achieve the optimal performance is obtained,which the regularization term must satisfy and is related with the size of images,the distribution of the real signal and the noise variance of observed data.Furthermore,it is proved that only when the noise variance is very small,can the lk norm constraint implement the optimal performance.It can provide the tutor for the design of iterative arithmetic.Simulation experiment demonstrates the validity of the conclusion.