Research On Video And Image Defogging Algorithm Based On Dark Channel Prior

With the rapid development of modern society, human beings have high density of the economic and social activities, which result in large emissions including automobile exhaust, the harmful gas from garbage incineration, industrial pollutants, and dust from construction sites and so on. In recent years, the haze days appear frequently, which seriously affects people's lives. Images and videos produced in the fog is not very clear. But in a lot of cases, there is a need to extract and analyze the detail information of the images and videos. Then some visual application systems, such as video surveillance system and navigation system, may lose their efficiencies. Therefore, the defog processing for the foggy images and videos becomes more and more important as the research hotspot, and it has theoretical and practical significance.This paper aims at the defogging processing research on the single image and video, including three aspects like theories of algorithms, simulation experiments, and the effect evaluation. The specific research work is as follows:Firstly, in terms of the defogging processing on single image, three algorithms including histogram equalization algorithm, the single-scale Retinex algorithm and the algorithm based on the dark channel prior are considered. These algorithms have been simulated and analyzed. What's more, we also compare the effect of various algorithms in the image processing. The effects of the algorithms are evaluated by the structural similarity, color reduction degree and detail intensity. Given that dark channel algorithm is proposed mainly for image defogging processing, its theory and method are more suitable in the haze removal than the other algorithms. However, the dark channel algorithm will lead to the over enhancement in the sky area, so in this paper we improve this algorithm on the atmosphere light and transmission map. In the end, the enhancement in the sky area will be natural.Secondly, we do research on the defogging processing of the foggy videos. Based on the theory that the video is composed of many image frames. If parameters such as atmospheric light, transmission map of each frame are calculated, it may lead to slow processing, and the discontinuous of video images. The background is consistent for the video produced by fixed camera. When using the method of background subtraction, we can apply the parameters of the first frame to the later video frames, which can avoid calculating parameter values repeatedly. This method reduce the amount of calculation, and can make the processed video more continuous.Finally, this paper integrates all algorithms. An experimental system which can make the fogging images and videos defogged is designed and implemented. Then the defogging result can display in the interface. Meanwhile, the system is able to calculate the algorithms' time consuming.In this paper, a lot of simulation experiments are carried out. After comparing the defogging effects of the different algorithms, we can summarize that the dark channel algorithm on single image defogging processing is reasonable and effective. The improved algorithm which is based on dark channel theory is put forward. The improved thoughts can be realized from two aspects. Block comparison method is adopted on the calculation of atmospheric optical value, and a compensation value is added to transmission map. According to the values obtained, the defogging effect is better in processing fogging images.