Retinex Based

Retinex-based methods aim to enhance image quality, particularly in low-light conditions, by separating an image into illumination and reflectance components, mimicking human visual perception. Current research focuses on integrating Retinex theory with deep learning architectures, such as transformers and state space models, to improve performance and address limitations of traditional approaches, including noise handling and detail preservation. These advancements are significantly impacting image processing applications, particularly in areas like low-light imaging and image fusion, by producing higher-quality images suitable for various computer vision tasks.