Fast single image dehazing based on color cube constraint

The outdoor images captured in bad weather are prone to yield low and poor visibility, which is a serious issue for most computer vision applications. The majorities of existing dehazing methods rely on the dark channel prior (DCP) assumption and therefore share two mains limitations; the model is invalid when the scene is intrinsically similar to the atmospheric light and the DCP method suffers from high computational cost to refine the transmission map. In this paper, we propose a fast single image dehazing based on color cube constraint based on new haze imaging model to overcome these two limitations. The thickness of the haze can be estimated effectively, and a haze-free image can be recovered by adopting the new method and the new haze imaging model. In this method, we first design a new haze imaging model which enables us to represent the hazy image inside a color cube according to the concentration of the haze. Then, to get an accurate value of the global atmospheric light we took the maximum value of each RGB color channel. Next, we propose a simple but very powerful prior or method called variation of distance prior (VOD), which is a statistic of extensive high resolution outdoor images. Using this prior combined with the designed haze imaging model and improved global atmospheric light, we can directly estimate the transmission map and restore a high quality outdoor haze-free image. The experimental results show that our model is physically valid, and the proposed method outperforms several state-of-the-art single image dehazing methods in terms of effectiveness robustness and speed.