To solve the random gauss noise in the video, this paper proposed a spatiotemporal combination method with the combing of background extraction and foreground filtering. Video images in this paper are divided into two parts: The foreground and the background, using background difference method combined with image expanding. Then deal with the two parts using spatiotemporal combination method based on the Nonlocal means filter and direct meaning in time domain. And then the output video can be got by adding the foreground and the background which are proposed. At last, this paper give the comparison of the Nonlocal means filter and the method in this paper. Experimental results indicate that the PSNR of the tow test video after denoising reaches 33.0043、29.0365 and 35.8340、31.5261. Experiment shows that for the monitoring video with fixed background, the arithmetic in this paper effectively deal with the lowfrequency information and the highfrequency details, on the basis of reducing the complexity and improving the realtime ability of the method.