Abstract: To optimize the objective function by using only the color and motion features of video,the consistent completion method was unable to predict the pixel value of unknown region accurately.In this paper,the structure information of video is introduced into the objective function.Firstly,calculate the space-time pyramid of damaged video.In the first layer of the pyramid from top to bottom,repair the pixels on the boundary of the broken part.The broken pixel can be inpainted by a weighted average method,update the boundary and inpaint the pixels on the new boundary by redefining weight,complete the first pyramid.Finally,the same method is used to inpaint each layer pyramid in turn,until all damaged pixels are completed.The various methods are used to repair the videos of different complex scenes and evaluate them with peak signal to noise ratio and structural similarity.Experimental results show that the proposed method achieves better visual effect,and the peak signal to noise ratio(PSNR)increases by 1 to 3 dB on average.