Beyond Brightening Low-light Images

Images captured under low-light conditions often suffer from (partially) poor visibility. Besides unsatisfactory lightings, multiple types of degradation, such as noise and color distortion due to the limited quality of cameras, hide in the dark. In other words, solely turning up the brightness of dark regions will inevitably amplify pollution. Thus, low-light image enhancement should not only brighten dark regions, but also remove hidden artifacts. To achieve the goal, this work builds a simple yet effective network, which, inspired by Retinex theory, decomposes images into two components. Following a divide-and-conquer principle, one component (illumination) is responsible for light adjustment, while the other (reflectance) for degradation removal. In such a way, the original space is decoupled into two smaller subspaces, expecting for better regularization/learning. It is worth noticing that our network is trained with paired images shot under different exposure conditions, instead of using any ground-truth reflectance and illumination information. Extensive experiments are conducted to demonstrate the efficacy of our design and its superiority over the state-of-the-art alternatives, especially in terms of the robustness against severe visual defects and the flexibility in adjusting light levels. Our code is made publicly available at https://github.com/zhangyhuaee/KinD_plus.