This is a page that curates AI-related papers published worldwide. All content here is summarized using Google Gemini and operated on a non-profit basis. Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.
This paper presents a novel approach for integrated image restoration, a critical task in low-level vision. Existing methods are either task-specific or rely on paired datasets for training, resulting in poor generalization performance and closed-set constraints. To address these issues, we propose a dataset-free, integrated approach utilizing recursive posterior probability sampling with a pretrained latent diffusion model. The method integrates a multimodal understanding model to provide semantic prior information to the generative model under task-independent conditions, uses lightweight modules to align degraded inputs with the generative preferences of the diffusion model, and employs recursive refinement for posterior probability sampling. Extensive experiments demonstrate that the proposed method outperforms state-of-the-art methods, validating its effectiveness and robustness. Code and data are available at https://github.com/AMAP-ML/LD-RPS .
