Coarse-to-Fine Global-Local Integration for Low-Light Image Enhancement

Low-light image enhancement (LLIE) aims to address issues such as insufficient brightness and loss of details in images captured in low-light environments. Existing LLIE methods generally focus only on global contrast adjustment or local detail enhancement, failing to effectively integrate global and local information, and are difficult to adapt to complex scenes with varying brightness distributions. To address these issues, we propose a coarse-to-fine global-local Integration network (GLIN) for low-light image enhancement. For global features, we introduce a self-attention mechanism to model long-range dependencies, and design a global feature enhancement module (GFEM) to effectively enhance image brightness while avoiding overexposure. For local features, we use a hybrid-scale feature fusion strategy to extract multi-scale detail features, designing a local feature enhancement module (LFEM) to improve edge sharpness and texture fidelity. Additionally, we design a progressive global-local fusion strategy that gradually integrates global brightness information and local detail features through multiple global-local integration modules (GLIM), achieving a coarse-to-fine multi-stage progressive enhancement. Quantitative and qualitative experimental results show that the proposed method effectively improves the image quality and visual perception in low-light scenes, and outperforms existing state-of-the-art methods.