Visibility is one of the important meteorological factors, and low visibility affects the safe and efficient operation of expressway. Visibility calculation method based on video data is similar to the perception of human eyes, and has the adaptability to bad weather conditions or complex observation sites. Based on three weather processes of fog in Anhui province on October 17th, December 1st and 7th of 2016, the visibility calculation model were built by using gray mean square error of corresponding video image and visibility observation data of three highway sites in Anhui province. Results show that the modeled visibility can well reflect the actual visibility corresponding with the change of fog process. The proposed visibility calculation method based on video image is similar to the perception of human eyes, i. e., the higher the visibility, the higher the resolution of terrain features for human eyes, and the greater the variance of the monitoring image as well. However, due to the difference in monitoring principle and the environment properties, error exists between the observation data and the calculated results. Theoretically, the error is low under the conditions of fog with uniform spatial distribution, and the visibility meter being close to the surveillance camera.