A low-light image enhancement method based on atmospheric scattering model

Since the traditional low-light image enhancement algorithm has poor generalization and is difficult to adapt to the complex and volatile real scenes, this paper proposes a low-light image enhancement method based on atmospheric scattering model. First, a low-light image is inverted for obtaining a foggy image. Subsequently, the inverted image is used as the input, which is passed through the atmospheric light estimation module and the transmission estimation module, respectively. Finally, the parameters obtained from this solution are incorporated into the atmospheric scattering model. Residual-in-residual dense blocks (RRDB) are introduced to extract high-level feature information from the inverted images. The dilated convolution is used in the atmospheric light estimation module with a larger receptive field and a minor number of parameters, thus the training speed is improved. A large number of experimental results demonstrate that our method achieves better quality compared with the state- of-the-art methods.