Low-light image enhancement based on Transformer and CNN architecture

Affected by low photon counts and noise, images captured in low-light environments commonly have low signal-to-noise ratios, details losing, low contrast and other artifacts. Low-light image enhancement which restores clear images from low-light images with unsatisfactory quality is challenging. In this paper, we propose a strong baseline model based on transformer and CNN architecture for low light image enhancement (TrCLLE). In TrCLLE, the advantages of transformer in effectively modeling long-distance dependence and convolutional neural network in modeling local features through inductive bias are integrated. TrCLLE consists of three parts: shallow feature extraction module, deep feature extraction module and high-quality image reconstruction module. At last, the factors affecting performance are analyzed in detail and the experimental results demonstrate that the proposed method achieves enhanced images with higher quality compared to other popular approaches.