Single Image Dehazing Using Sky Adaptive Fusion

Abstract

Hazy days often appears in our daily life, which reduces the visibility of images captured from various cameras. High quality dehazing algorithms are in great demand. Most of the common approaches are based on the dark channel prior (DCP). However, DCP is not valid in the sky region and make the estimated transmission close to zero, which lead to color distortions and noise appear in the sky region. To deal with the above problem, this paper proposes a bright channel prior (BCP) to for computing the transmission map in the sky region. After that, the transmission maps obtained using DCP and BCP are fused effectively using the proposed gradient-based fusion scheme. The haze-free images are finally recovered using the fused transmission map and input hazy image. Experimental results demonstrate that the proposed method outperforms the DCP method especially on preserving the color consistence of sky regions.