Improved single image dehazing using dark channel prior
Yan Wang, Bo Wu · 2010
Atmospheric conditions induced by suspended particles, such as fog and haze, severely degrade image quality. Haze removal from a single image of a weather-degraded scene remains a challenging task, because the haze is dependent on the unknown depth information. In this paper, we introduce an improved single image dehazing algorithm which based on the atmospheric scattering physics-based models. We apply the local dark channel prior on selected region to estimate the atmospheric light, and obtain more accurate result. Experiments on real images validate our approach.