水平路径上多帧湍流退化图像重建

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.

Horizontal-path turbulence degraded image multiframe reconstruction

In order to restore degraded images through horizontal-path turbulence, a multiframe 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.