A moving objects based real-time defogging method for traffic monitoring videos

In this paper, a moving objects based real-time defogging method for traffic monitoring videos is proposed. Firstly, dark channel prior based image defogging method has been improved. Then, the proposed image defogging method is used for traffic monitoring video defogging. To improve the processing speed, the correlation between the adjacent frames of videos is exploited. The moving objects are detected using adjacent frame difference method. The frame content is divided into moving foreground and background. Afterwards, the foreground and background are processed with different defogging manners to reduce the computational complexity of defogging processing. Experimental results show that the proposed method can generate a good defogging effect which will facilitate the subsequent intelligent traffic analysis. Furthermore, the proposed method is fast enough to process the standard-definition videos at the speed of 26 frames per second on average.