Research On Methods Of Fog Removal In Surveillance Situations

In foggy conditions, due to the low visibility of the scene, the video or image obtained by imaging device capturing visible light will be severely degraded, such as contrast reduction, color attenuation, etc. This will affect the normal work of the outdoor computer vision applications system, such as video surveillance, forest fire monitoring, etc. In order to provide real-time high quality images for image processing applications, many efforts have been placed on fast single image fog removal according to the characteristics of video surveillance. This paper gives a deep research in fog removal from two aspects based on the atmospheric scattering model in fog days, including image enhancement and image restoration.First of all, the atmospheric scattering model in fog days is studied in this paper, and specific analysis of imaging process is given.Secondly, some research has been given in Histogram Equalization and Homomorphic filtering algorithm from the perspective of the enhancement foggy image. Combined with the advantages and disadvantages of the two algorithms, the single image fog removal algorithm has been proposed based on the Contrast Limited Adaptive Histogram Equalization and Homomorphic filtering algorithm. Experiments and analysis show that the algorithm improved to some extent, the quality of foggy images.Then an improved algorithm named fast image fog removal based on gray image Guided filtering is proposed on the basis of the Dark Channel Prior. First, estimate atmospheric scattering model transfer chart through the dark channel information, and then use the gray image of the input foggy image to guide the transmission image filtering, on the basis, the recovery image with no fog will be obtained. Experiments demonstrate that our algorithm can effectively remove fog from a foggy image while increasing the efficiency.Finally, in order to apply the proposed algorithm to the engineering practice, the algorithm has been implemented through C/C++language, using the efficiency of C language and some programming methods to further improve the efficiency of the algorithm. The results show that the algorithm can achieve real-time requirements of monitoring scenarios, and can get a satisfactory outcome of fog removal.