Nonhomogeneous image dehazing based on adaptive scattering coefficient

• 1、School of Electronic Information and Artificial Intelligence, Shaanxi University of Science & Technology, Xi’an 710021, China

Abstract：Aiming at the problems of color distortion or too deep and dehazing residue in the traditional method of removing nonhomogeneous haze, a dehazing method for nonhomogeneous images based on adaptive scattering coefficient is proposed. Firstly, three kinds of haze density features are used to characterize the nonhomogeneous distribution of haze, innovatively establishes the positive correlation function between the atmospheric scattering coefficient and the haze concentration, it realizes the dynamic identification of the haze concentration of different pixels, and adaptively allocates different atmospheric scattering coefficient values according to different haze concentrations at each pixel, and obtain more accurate transmittance values. Secondly, the context regularization based on Robinson\'s high-order operator is used to constrain the transmittance, avoiding halo artifacts in the image after dehazing, and retaining image texture and detail information to the maximum extent. The experimental results show that the proposed method has better results in visual effects and objective indicators on real haze images and nonhomogeneous haze images.

Keywords： adaptive scattering coefficient nonhomogeneous haze characteristics of haze concentration weighted fusion context regularization