A Novel Transformer-Based Approach to Low-Light Image Enhancement

Image quality significantly impacts the effectiveness of computer vision (CV) applications such as autonomous driving, facial recognition, and remote sensing. However, inadequate lighting substantially degrades image quality, resulting in muted colors, reduced visibility, distortion, increased noise, and loss of detail, which can significantly interfere with the performance of these CV-based applications. In response, our research introduces the Fusion Low-Light Former (FLLF) to address the critical need for enhancing images captured in low-light conditions. Our method uses transformer-based techniques and feature fusion to effectively merge low-and-high-level features in degraded images in a hierarchical architecture. Consequently, our FLLF markedly improves low-light image enhancement performance.