基于Poisson模型的湍流退化图像多帧迭代复原算法

Processing the measurement delay INS/SAR integrated navigation in-coordinate interval filtering algorithm study

随着航天空间技术的发展,空中目标的成像探测研究越来越重要。受大气湍流的干扰,观测到的目标图像是严重模糊的。为了从观察到的多帧含噪的湍流退化图像中将目标原图像有效地恢复出来,本文提出了一种新颖的基于图像统计模型的图像复原算法。跟据图像Poisson概率模型建立了有关多帧图像数据的对数似然函数,通过极大化该对数似然函数,推导出了目标图像及各帧图像点扩展函数的迭代求解关系。同时,将点扩展函数的支持域等先验条件有效地加入到迭代计算过程中,以便快速地利用迭代技术将目标图像和各帧点扩展函数估计出来。该算法能用少数帧图像极大程度地恢复出目标图像。为了验证本文算法的恢复效果和可靠性,对在强噪声污染条件下的湍流退化图像进行了恢复实验,实验结果表明本文算法对空中目标湍流退化图像的复原是非常有效的。

With the development of astronomical space technology, the study of imaging detection of space object becomes more and more important. The observed object images are severely blurred because of the influence of atmospheric turbulence. A novel restoration algorithm based on image statistic model is presented in this paper for restoring object image from a sequence of turbulence-degraded images with noise. The logarithmic maximum-likelihood function of multi-frame images is built according to the image Poisson models, and the iterative relationships to estimate the object image and PSFs are derived by maximizing the logarithmic maximum-likelihood function. Meanwhile, a priori knowledge about the support of the PSFs is incorporated into the iterative process of the restoration as that the object image and the PSFs can be quickly estimate from the turbulence-degraded images. The proposed algorithm can restore the object image as completely as possible by using a few of frames images. In order to test the validity of the proposed algorithms series of restoration experiments are performed on the turbulence-degraded images with heavy noise and the experiments results show that the proposed algorithm is effective to restore the space object from their turbulence-degraded images.