Abstract: To solve the problem of motion blur caused by relative movement between camera and target during the shooting process, based on the gradient cepstrum, least-square algorithm and differential auto-correlation, an algorithm was proposed to estimate parameters of the point spread function (PSF) of motion blurred image. The blur direction was estimated with the method of the gradient cepstrum analysis and least-square method, and then the motion length was estimated with the differential auto-correlation method. PSF parameters of the blurred simulation and real images were estimated with the proposed algorithm, when the angle of blurred direction was in the range of 0°~179°,blurred scale was in the range of 5~60 pixels, the absolute error of blurred angle was less than 2°, and the absolute error of blurred length was less than 2 pixel by the algorithm estimation. Compared with other literature algorithms, the results show that the motion blur parameters can be accurately identified with the proposed algorithm.