A novel haze image enhancement algorithm

Abstract

The atmospheric scattering model (ASM) is usually used in the haze image enhancement. Researchers always assume that the scene incident light is fixed, which is unreasonable in the real environment. In this paper, an improved atmospheric scattering model (IASM) with the incident light redefined is proposed to enhance the haze images. Firstly, the bright channel map and fuzzy clustering algorithm are used to divide the haze image into different scenes and estimate the incident light of each scene. Secondly, the atmospheric light is estimated based on the sky area location and the transmittance. Finally, the albedo of the image is restored through the IASM. The experiments show that the proposed algorithm is effective in enhancing the haze image.