<p>Aiming at the problem of defogging aerial photography scenes with highly coupled light and fog distributions, traditional methods have problems such as distortion of global assumptions of atmospheric light, deviation of linear modeling of transmittance, and failure of sky region correction. In this study, we propose a defogging method for aerial images that integrates the variational retinex theory and fog transmittance model. First, a spatially variable atmospheric light intensity model is constructed to accurately portray the light attenuation characteristics. Second, the luminance component after the guided filtering process is used as the initial iteration condition of the efficient variational retinex to obtain the atmospheric light and reflection maps, and the texture details of the reflection map are enhanced by gamma correction. Third, the physical characteristics of the fog are analyzed to establish a nonlinear mapping model of the fog concentration and transmittance of the scene. Fourth, a sky region adaptive correction strategy is introduced to solve the transmittance distortion problem. Finally, the optimal fog-free image is reconstructed using the improved atmospheric scattering model. The experimental results show that the proposed algorithm performs well in both subjective and objective evaluations.</p>