Abstract
Image restoration involves recovering a high-quality clean image from itsdegraded version. Deep learning-based methods have significantly improved imagerestoration performance, however, they have limited generalization ability todifferent degradation types and levels. This restricts their real-worldapplication since it requires training individual models for each specificdegradation and knowing the input degradation type to apply the relevant model.We present a prompt-based learning approach, PromptIR, for All-In-One imagerestoration that can effectively restore images from various types and levelsof degradation. In particular, our method uses prompts to encodedegradation-specific information, which is then used to dynamically guide therestoration network. This allows our method to generalize to differentdegradation types and levels, while still achieving state-of-the-art results onimage denoising, deraining, and dehazing. Overall, PromptIR offers a genericand efficient plugin module with few lightweight prompts that can be used torestore images of various types and levels of degradation with no priorinformation on the corruptions present in the image. Our code and pretrainedmodels are available here: https://github.com/va1shn9v/PromptIR