Haze Removal Using Color Attenuation Prior

Foggy images and video in the environment,there are poor resolution,low contrast,color distortion and other serious degradation.It directly affect and limit the role of outdoor visual system.Therefore,in order to effectively improve the fog-degraded image and video quality,reduce the effect of the fog,haze weather to outdoor vision systems,there are important practical significance for haze removal to images and videos effectively.This paper mainly studies the algorithms of haze removal to images ans videos,the research content is as follows:(1)Aiming at the shortcoming of the image haze removal algorithm using color attenuation prior,mainly from the scene depth estimation and atmospheric scattering coefficient estimation two aspects to improve.Firstly introduce the improvement about scene depth estimation,by establishing a nonlinear model of scene depth and learning the parameters of the model with training sample,it gets more accurate scene depth.Secondly introduce the improvement about atmospheric scattering coefficient estimation,by defining the function of atmospheric scattering coefficient and getting the parameters of the function with a large number of image,it gets more accurate atmospheric scattering coefficient.(2)Aiming at the shortcoming of the existing video haze removal algorithm,the each frame of defog video is not harmonious in color and structure,it has a large time consuming.In this paper,it presents the video haze removal based on universal atmospheric optical value and universal transmission.This method takes advantage of surveillance video characteristic,universal atmospheric optical value and universal transmission are applied to the each frame of fog video.Firstly extracting the background image from the fog video,then obtaining the atmospheric optical value and transmission of background image,finally the atmospheric optical value and transmission are applied to the each frame of fog video,realize video haze removal.(3)Comparing the proposed defog method to haze removal using dark channel prior and haze removal using color attenuation prior,from the qualitative and quantitative two aspect analysis the superiority of the proposed defog method.And by the experiment shows that the proposed universal atmospheric optical value and universal transmission is effective to fog video.