采用基于单步预测的卡尔曼滤波器,利用含噪声图像的自身特性估计噪声参数,研究了一种对多帧静止相干激光雷达距离像进行复原的方法。通过对仿真图像进行复原,研究了在可利用图像帧数较少时的滤波性能和帧数足够多时滤波性能的极限值,并与在相同帧数下的图像平均法进行比较。结果表明,该算法收敛速度快、具有更高的图像改善信噪比和极限值。 A Kalman filter based on single-step prediction is used to estimate the noise parameters based on the noise-free images. A method to reconstruct the distance image of multi-frame stationary coherent laser radar is studied. By restoring the simulation image, the limit of the filtering performance when the number of available image frames is small and the filtering performance is enough are studied and compared with the image averaging method with the same number of frames. The results show that the proposed algorithm has the advantages of fast convergence, higher image quality and improved signal-to-noise ratio and limit value.