Aiming at the problems that the existing iterative blind image restoration algorithms are mostly time-consuming and their convergence guaranteeing is difficult, a fast algorithm was proposed. Firstly, the spectrum of original image is reconstructed in accordance with the power law. Then, utilizing the relationship between the spectrums of the orignal and reconstructed images, the Point Spread Function (PSF) is estimated by the multidirectional comprehensive estimation, which can reduce the estimation error and enhance the stablility. Finnaly, the restored image is obtained by using the estimated PSF and Wiener filter. The proposed algorithm was compared with the existing algorithms by experiments. The results show that for the class G PSFs which fit for the numerous imaging systems, the proposed algorithm can obtain more accurate PSF, and reduce the ringing artifacts, yielding restored images with higher quality.