Single Image De-haze under Non-uniform Illumination Using Bright Channel Prior

Recent single image de-haze approaches assume the atmospheric light is the only illumination in one haze image and use a globally constant to image de-haze. However, every local pixels in an outdoor image is actually under the influence of non-uniform illumination in real world. The accuracy of the environmental illumination estimation has a great influence on the result, so the traditional haze image model is defective and not conforms to the facts. In this paper, we refine the haze image model under the consideration of nonuniform illumination, and then we propose a new prior called “bright channel prior” to de-haze single image combining with the dark channel prior. The bright channel prior, which inspired by the dark channel prior [1], is a statistic of haze-free outdoor images. Using this prior, we derive a more accurate environmental illumination estimating algorithm for single image de-haze. Our experiments prove the feasibility of the method we propose and outperform other image haze removal approaches.