Retinex Decomposition

Retinex decomposition is a computational image processing technique that separates an image into illumination and reflectance components, aiming to improve image quality, particularly in low-light conditions. Current research focuses on integrating Retinex with deep learning models, such as diffusion models and convolutional neural networks, often within a plug-and-play framework, to enhance denoising and color correction capabilities. These advancements leverage the physical interpretability of Retinex while harnessing the power of deep learning for superior performance in tasks like low-light image enhancement and other vision applications, leading to improved results compared to traditional methods.