运动模糊图像的Kalman滤波盲复原研究

Research of Motion Blurred Image Restoration by Kalman Filtering

本文提出一种对由于模糊参数未知的运动模糊和随机噪声引起降质的图像进行复原的方法。对于一幅这样的图像,首先确定图像退化过程的参数,即其点扩展函数(PSF);再假设图像可由一个半因果的随机场表示,则图像表示和图像退化模型可以写成矩阵-向量形式。然后将它们分别作为状态方程和量测方程可以推导出N个频域中的并行Kalman滤波器。实验结果表明这种结合PSF估计和Kalman滤波复原的图像处理方法效果是令人满意的。

A restoration method for images degraded by both unknown parameter motion blur and random noise interferenee is proposed in this paper. To begin with the image restoration, parameters of the imaging system are identified, then its point-sprad function (PSF) is obtained; it is assumed that the image can be represented by a semicausal random field, then both the image representation model and observation model can be written in matrix-vector formulations. N parallel Kalman filters in the frequency domain are derived based on these formations. Experiment shows the effect of this method, which compounded PSF estimation and restoration by Kalman filtering is satisfying.