A Variational Retinex Model With Structure-Awareness Regularization for Single-Image Low-Light Enhancement

Low-light image enhancement (LLIE) is a method of improving the visual quality images captured in weak illumination conditions. In such conditions, tend to be noisy, hazy, and have low contrast, making them difficult distinguish details. LLIE techniques many practical applications various fields, including surveillance, astronomy, medical imaging, consumer photography. The total variational sound solution this field. However, requirement an overall spatial smoothness map leads failure recovering intricate This paper proposes that interaction between global detail recovery Retinex model can optimized by adopting structure-awareness regularization term. resultant non-linear more effective than original one for LLIE. As model-based method, its performance does not rely on architecture engineering, super-parameter tuning, or specific training dataset. Experiments proposed formulation challenging low-light yield promising results. It shown only produces visually pleasing pictures, but it also quantitatively superior calculated full-reference, no-reference, semantic metrics are beyond most state-of-the-art methods. has better generalization capability stability learning-based Due flexibility effectiveness, deployed as pre-processing subroutine high-level computer vision applications.