Inland waterway image haze-removal based on the dark-channel prior

Abstract

Dark-channel prior is a perfect theory to restore the outdoor haze image, but the color distortion is emerged when processing the bright region such as sky. Especially for inland river haze image, there is a large area of sky in image. To solve this problem, an improved dehazing algorithm is proposed in this paper. It first segments the haze image into sky and non-sky region through image segmentation algorithm, and then the dark-channel prior is used to deal with two regions to obtain the adaptive estimation of the transmission rate respectively. Finally, two transmission rates are combined to restore the inland river haze image. Experimental results show that the proposed algorithm provides significant advantage in producing more natural sky region, while achieving the image restoration of the non-sky region in inland river haze image.