Single image de-haze based on a new dark channel estimation method

The image quality is effected dramatically by the weather conditions such as haze, fog and smoke, in surveillance or vehicle systems. In this paper, we propose a novel algorithm to remove haze from a single image. A new dark channel estimation method is employed by subtracting three times variance from mean to approximate the minimum value, which is defined as the dark channel in a local area. To achieve anisotropic local mean and variance, a fast bilateral filter is introduced. Its complexity is a linear function of the size of the image, which makes real-time visibility restoration possible. Compared with existing algorithm based on dark channel prior, the proposed algorithm have better performance by reducing the halos introduced by large patches very well with less computational cost.