The resolution of space object images was decreased seriously by the atmospheric turbulence, photon noise and collimating fault of optical track system. The multiframe blind deconvolution algorithm for improving the resolution of object images based on the maximum-likelihood estimation was presented. From the record images the object function and point spread function were estimated by the conjugation gradient optimization method. The reduction of photon noise was reached by using the smoothing low-pass filter during the algorithm iteration. The restoration results of simulation and real data show that the proposed blind deconvolution algorithm can overcome the effects of the atmospheric turbulence, photon noise and collimating fault of optical system on object images, and improve the resolution of object images to arrive at the level of diffraction-limited.