Image Dehazing Based on Accurate Estimation of Transmission in the Atmospheric Scattering Model

Image dehazing is a challenging and highly desired technology in computer vision applications. The dark channel prior (DCP) has been considered to be an efficient dehazing technique in recent years. However, the invalidation of DCP can induce unreliable estimation of transmission, resulting in inaccurate color information recovery, halo artifacts, and block effect. In this paper, a novel brightness map is proposed based on the observation on outdoor haze-free/haze images that can reflect the brightness information and the light reflection ability of the scene, furthermore, the relationship between DCP and the brightness map is given in mathematical model. The proposed algorithm can compensate for the DCP effectively, estimate the transmission map accurately, get the global atmospheric light adaptively and segment the image automatically. Using multiscale guided filter refine transmission map, the halo artifacts are able to be avoided in the scene depth of a sudden change. A series of experiments are additionally implemented to demonstrate that the proposed algorithm can obtain high-quality haze-free images with abundant distinguished details, low color distortion, and little halo artifacts that can outperform or be comparable with four state-of-the-art haze removal algorithms.