Research On Single Haze Image Restoration Under Non-uniform Illumination

The turbid medium is full of atmosphere, in haze weather. Outdoor haze image is degraded by absorbed or scattered because of the turbid particles. Currently, image de-haze approaches under uniform illumination in the daytime that the atmosphere light is mainly from the sky made significant progress at home and abroad. However, the research of image de-haze approaches under non-uniform illumination where exists artificial light is poor. Different from the imaging condition in uniform illumination, the haze images existing artificial light is more difficult to deal with. They suffer from both unbalanced illumination and haze. Therefore, in this paper we research image dehazing under non-uniform illumination deeply.In this paper, we proposed a novel efficient dehazing method under non-uniform illumination for nighttime haze condition based on haze image model and dark channel prior theory firstly. Then the proposed algorithm is popularized that suits both haze images in the nighttime and haze images under non-uniform illumination in the daytime. Our algorithms can achieve both illumination balanced and haze free results. This thesis research includes the follows contents:(1) Nighttime haze image. Firstly, a detecting algorithm is proposed to confirm the position of artificial light sources; Secondly, atmospheric light and compensatory coefficient are obtained according to artificial light characteristic; After that, the haze-free image is effectively recovered based on a newly proposed imaging model; Finally, edge based color constancy is used to correct color. Experimental results show that our algorithm can get good performance under non-uniform illumination.(2) Daytime haze image contains non-uniform artificial light and uniform air light. We popularize the algorithm so that they suits both haze images in the nighttime and haze images in the daytime. Firstly, we remove artificial light based on detection of artificial light sources and estimation of artificial light. Secondly, we estimate transmission and refine transmission based on CLMF after removing the influence of artificial light. Finally we get haze-free image.