Motion blurred images have a wide application in film and television and computer visual technology. The paper, aimed at solving the restoration problems of motion blurred images, regards the mapping from dynamic blurred images to sharp images as a kind of mode mapping, and proposes a motion blurred image restoration algorithm based on GAN in combination with convolution. By learning the mode via the training of image sets, the fitting output can achieve that sharp images correspond to the blurred images. As is demonstrated by the experimental results, the image restoration effect by the algorithm is remarkable and the definition is high.