Degradation Aware Visual Prompt

Degradation-aware visual prompts are emerging as a powerful technique in image restoration, aiming to improve the accuracy and versatility of models handling diverse image degradations (e.g., noise, blur, weather effects). Current research focuses on integrating these prompts into diffusion models and vision transformers, often leveraging pre-trained large language models to generate semantically rich and degradation-specific prompts that guide the restoration process. This approach offers a promising path towards universal image restoration models, capable of handling multiple degradation types with a single architecture, reducing storage costs and improving flexibility compared to task-specific methods.