Single-Image Dehazing Based on Improved Bright Channel Prior and Dark Channel Prior

Single-image dehazing plays a significant preprocessing role in machine vision tasks. As the dark-channel-prior method will fail sky region of image, resulting inaccurately estimated parameters, and given failure many methods to address large band haze, we propose simple yet effective for single-image based on an improved bright prior dark channel prior. First, use Otsu by particle swarm optimization divide hazy image into regions non-sky regions. Then, estimate parameters physical model. Second, weighted fusion function efficiently fuse two priors. Finally, clear is restored through Experiments illustrate that our can solve problem invalidation well achieve high-quality restoration, especially images with limited haze.