Enhancing Low-light Images Using a Coarse-to-fine Linear Mapping Strategy

Conventional low-light image enhancement methods using linear mapping are simple but often result in poor visual quality due to their limited capability. While deep learning techniques have improved these outcomes, they usually depend on complex prior models, such as illumination and noise estimation. To address these challenges, we propose a new method that combines deep learning with traditional linear models, utilizing a coarse-to-fine linear mapping strategy. This approach avoids prior modeling and effectively tackles image degradation. Our experiments show that our model, guided by this strategy, is both simple and effective, and it competes well with existing deep-learning methods for low-light image enhancement.