To improve the usability of a foggy video, an improved video defogging method based on fog theory was proposed. By using dark channel prior knowledge, Retinex method and image fusion, the method applies the values of global atmospheric light and a medium transmission map estimated from the video backgrfound image to defogging of all the video frames. The effects of the defogging for the video image were evaluated by three methods in subjective qualitative evaluation, objective quantitative evaluation and operation speeds. Experimental results demonstrate that the proposed method runs at 5.45 frame/s for a video image of 480×640, and it not only obtains a fast processing speed but also effectively avoids color jump during the process of restoring image. As the modified method uses the interval estimation to estimate the value of global atmospheric light, and combinates image restoration and image enhancement to obtain the value of medium transmission map, it improves the visibility and contrast of restored video image effectively as well as color effect as compared with the traditional video defogging methods.