Latent Degradation
Latent degradation research focuses on modeling and mitigating the unseen, underlying processes that corrupt data, such as compression artifacts in videos or rain streaks in images. Current efforts concentrate on developing deep learning models, often incorporating autoencoders, diffusion models, and transformer architectures, to explicitly learn and represent these latent degradations, enabling more effective restoration or enhancement. This work is significant because accurately modeling these hidden degradations improves the performance of various image and video processing tasks, leading to higher-quality results in applications ranging from medical imaging to remote sensing.
Papers
May 10, 2024
September 9, 2023
August 28, 2023
July 27, 2023
November 13, 2022