A motion detection and image restoration method based on video reconstruction from a single coded exposure photograph is proposed. The exposure time of each pixel in the sampling regions is modulated respectively, and a serial of video frames are reconstructed from the coded exposure image of each region. By using the frames of the videos, displacement vectors and point spread function are calculated, and used to restore the blurred image finally. The proposed approach needs no additional devices and has advantages in both systematic complexity and cost. The effectiveness of the proposed method is experimentally confirmed by the results of simulation on problems of motion detection and image restoration.