Research Of Dehazing By Retinex And Dark Channel Prior

Haze, fog is common natural phenomenon in the human life. The atmospheric water vapor is condensed into drop by the dust, and then it becomes haze. Haze has a bad influence to the photography and sensing system. On one hand, the surface reflection light occur attenuator and forward scattering due to the atmospheric particles scattering. On the other hand, the atmospheric light involve in imaging process because of the atmosphere scattering. Haze has adverse effects for the sensing system which has been widely used in national defense, automatic control, artificial intelligence and other fields. Therefore, the research of dehazing has a wide range of practical significance.In this paper, the research is to recover the haze-free image with the dehazing technology and soft matting technology. It mainly has these parts.First, this paper summarizes the existing dehazing method and have an introduction of the Atmospheric Model.Secend, this paper makes a dehazing result by the Retinex method.Third, this paper finds a shortage of the Dark Channel Prior.Forth, this paper improves the dark channel prior by the Learning Based Matting and have a good result.The most important feature of the research is to propose an idea which use the learning based matting to improve the dark channel dehazing method. The method of this paper not only retains the advantages of the dark channel dehazing, but also has an improvement in the noise problem and timeliness.