MEGF-Net: Low-Light Image Enhancement via Adaptive Multi-Exposure Adjustment and Fusion

Low-light image enhancement (LLIE) aims to improve visual quality and interpretability under low-light conditions. Images captured in such environments often exhibit low contrast, limited visibility and pronounced noise. Most existing methods rely on a single illumination estimation or mapping strategy, which limits their adaptability to complex low-light scenes and hinders effective balancing of detail restoration, high dynamic range adjustment and model efficiency. To overcome these challenges, we propose a lightweight and efficient Multi-Gamma Exposure Generation and Fusion Network, referred to as MEGF-Net. This method introduces a multi-exposure generation and fusion mechanism that jointly optimizes exposure regulation, detail recovery and colour consistency. Specifically, the Adaptive Multi-Exposure Adjustment (AMEA) Module predicts gamma coefficients, linear modulation factors and compensation parameters, and generates a set of candidate images with diverse brightness, contrast and colour responses, providing rich exposure representations and colour priors for low-light enhancement. The Multi-Exposure Fusion (MEF) Module then performs adaptive fusion using a channel-wise strategy and incorporates a perceptually guided compensation mechanism, which enhances texture fidelity and colour accuracy. Finally, the Color Refinement Module (CRM) calibrates colour consistency and refines structural details, mitigating artefacts and colour shifts caused by exposure differences and fusion errors. Extensive experiments show that MEGF-Net achieves superior enhancement quality on multiple public datasets compared with state-of-the-art methods while substantially reducing model parameters and computational cost. These results demonstrate the effectiveness, efficiency and practical potential of the proposed approach for LLIE. Our code is available at: https://github.com/98Hao/MEGF-Net.git.