MLSwinTNet: A Multi-Level Feature Interaction Network for Low-Light Image Enhancement

Abstract

Low-Light Image Enhancement (LLIE) is crucial for improving image quality and visual analysis. This study proposes MLSwinTNet, an LLIE network based on the Swin Transformer. The core MLSwinT module adopts a UNet-like design, integrating a dual-branch feature extraction module and a multi-level feature interaction module. The former combines CNN and Swin Transformer to efficiently capture both local and global features, while the latter optimizes detail and color restoration by adjusting and fusing feature levels. Experimental results demonstrate that MLSwinTNet outperforms state-of-the-art methods on standard datasets, achieving a PSNR gain of 27.33 dB and an SSIM score of over 0.96, providing a more competitive solution for LLIE.