水平路径上多帧湍流退化图像重建  

Horizontal-path turbulence degraded image multiframe reconstruction

作　　者：邵慧[1] 吴东升[1] 陈杰[1] 赵阳[1] 金海红[1] 周春[1] 

机构地区：[1]安徽建筑大学电子与信息工程学院,安徽合肥230601

出　　处：《电视技术》2017年第3期81-87,共7页Video Engineering

基　　金：安徽建筑大学博士启动基金项目(2015QD07);安徽省教育厅重点项目(KJ2016ZD149;KJ2015ZD14);国家自然科学基金项目(614710022)

摘　　要：为了重建水平路径上远距离成像系统获取的湍流退化图像,提出去除形变和振铃的分块多帧盲反卷积方法。首先按采集的顺序将图像序列分成若干图像组,通过B-spline基函数配准方法去除非等晕块之间的形变,第二步通过基于块的时域融合算法去除等晕块间的振铃,得到衍射受限的湍流退化图像,最后将所得的多帧图像进行频域重建,复原高质量图像。实验表明,所提出的算法适合于水平路径上获取的退化图像的重建,得到较高的主观视觉图像质量。In order to restore degraded images through horizontal-path turbulence, a muhiframe iterative blind deconvolution （MIBD） algorithm suit limitless isoplanatic patch size is proposed. Firstly, the video is divided into several groups of picture, remove image geometric deformation through B-spline based registration. Then eliminate ringing artifacts in the boundary regions through temporal fusion method which preserve the signal and suppress noise in the high-frequency area. Finally, the image and PSFs are obtained by multiframe iterative recursive projection scaled-gradient algorithm in frequency domain. Experiments using real data illustrate that this approach can effectively alleviate horizontal-path atmospheric turbulence, recover details of the scene and significantly improve visual quality.

关 键 词：大气光学 迭代盲反卷积 B-spline基函数 Tichonov正则化 时域融合滤波器 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]