Research On Video Defogging Algorithm For Road Traffic

Under haze or fog weather conditions, low road visibility might lead to blurredand lower contrast of the images obtained by traffic video surveillance system, it isdifficult to provide accurate image information for traffic supervision. Video andimages defogging technology can significantly improve the quality degradation ofimages caused by haze and fog, make the video obtained by surveillance systemrecover resolution, which can effectively assist traffic monitoring. Article mainlyaims at the road traffic video that affected by haze and fog, substantially eliminatingimage degradation and restoring the color and contrast of the scene. To get clearvideo, a method combining image restoration with image enhancement is used todeal with fog images according to the influence mechanism of fog and haze on pictureimaging.The main research content of this paper is to design a complete algorithm forvideo defogging that can be applied to road traffic scene based on single imagedefogging, and use foggy road traffic video for verification. Studying the fog andhaze influence mechanism on the video image, and deriving the image restorationmodel on the basis of the classic model of atmospheric light scattering. The darkchannel of foggy image is used to calculate transmittance. In order to speed up thealgorithm, take the pixel in foggy image that is the brightest pixel in dark channelmap as the estimates of atmospheric light. And the guided filter is used to refine thetransmittance. Based on image restoration model, foggy image can be restored.Besides, an image enhancement method based on constant hue is use to increaseimage’s brightness and contrast for image post-processing.Due to the information redundancy between frame images of video, a keyframe is selected every8-12frame,the atmospheric optical parameter of the keyframes is passed to the non-key frame image. And the pyramid LK optical flowmethod is using to calculate the motion vector of the non-key frame to obtain thetransmittance. Because pixel motion occlusion and other problems in video image,to achieve clear video and improved defogging effect the guide filter is used to corrected the error of transmittance map that is obtained by motion vectorestimation.The traffic video defogging algorithm based on the dark channel is simple andeasy to calculate, and can make fog and haze video clear. Through a lot ofsimulation experiments, the approach can achieve a good road traffic scenes videodefogging effect. Traffic video defogging algorithm can provide clear video imagesfor traffic supervision department, also can lay the theoretical and technical basis forthe research of intelligent traffic monitoring, vehicle video navigation, driverassistance and other aspects in future.