BEENet: A Balanced Enhancement Network for Uneven Exposure Low-Light Image Enhancement

To simulate the complex lighting variations in real-world low-light scenarios, we built a low-light image dataset with uneven exposure. Various low-light image enhancement algorithms were evaluated on this dataset, highlighting the limitations of existing methods in terms of color restoration and brightness balance in similar conditions. We proposed an improved model that optimizes detail retention and color consistency in unevenly exposed scenes. The effectiveness of the model has been validated through quantitative evaluation of experimental results and qualitative analysis of visual effects, demonstrating improvements in detail preservation, color consistency, and visual quality.