基于MAP估计的自适应光学图像非对称多帧盲复原  

MAP Based Asymmetric Multiframe Blind Restoration for Adaptive Optics Images

作　　者：杨阿锋[1] 鲁敏[2] 滕书华[2] 孙即祥[2] 

机构地区：[1]国防科学技术大学指挥军官基础教育学院,长沙410073 [2]国防科学技术大学电子科学与工程学院,长沙410073

出　　处：《计算机辅助设计与图形学学报》2014年第6期861-869,共9页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(60972114);中国博士后科学基金(2012M512168)

摘　　要：为了提高地基望远镜观测空间目标的分辨率,提出一种针对自适应光学图像的盲复原方法.利用多帧图像的盲解卷积作为高清晰复原的主要技术手段,在贝叶斯最大后验概率估计框架下推导出泊松噪声模型的通用代价函数;提出基于非线性方程寻优的代价函数求解方法,并将非对称迭代引入求解过程,避免求解陷入局部极值点,以保证复原结果的可靠性.实验结果表明,该方法对泊松噪声有较好的适应能力,能够有效地从多帧湍流降质图像中恢复出目标图像.In order to improve the resolution of space object images observed by ground-based telescope,a novel blind restoration method for adaptive optics images is presented.This paper uses the multiframe blind deconvolution as the main technique means for high resolution restoration,and proposes a general cost function to deconvolve Poisson noise model image under the framework of Bayesian maximum a posteriori.To minimize the cost function,two algorithms based on nonlinear optimization method are proposed.In addition,asymmetric iteration method is adopted to avoid converging to local minima and maintain the robustness of restored images.Experimental results show that the proposed method can effectively recover high quality image from turbulence degraded images and alleviate the negative influence of noise.

关 键 词：图像复原 盲解卷积 最大后验概率 交替迭代 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]