Reversed and Fused Zero-DCE

Zero-DCE is a deep learning-based low-light enhancement algorithm. The paper researches the network structure of the Zero-DCE algorithm and proposes an exposure correction algorithm based on Zero-DCE, fuses this exposure correction algorithm and the original Zero-DCE algorithm and proposes a network capable of performing both low-light enhancement and exposure correction. First, the study of the Zero-DCE algorithm found that the key to its low-light enhancement lies in the parameter map generated by the DCE-Net, so the parameter map was modified, and the algorithm has the function of exposure correction at this time. Secondly, the transformed parameter map and the original parameter map are fused, and then the algorithm has the functions of low-light enhancement and exposure correction simultaneously. Based on Zero-DCE, the paper proposes an image enhancement method without reference pictures, which can realize low-light enhancement and exposure correction of images simultaneously, and at the same time, an exposure correction method is provided.