Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2509.24580 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916976562012160 |
|---|---|
| author | Wang, Lingyu Meng, Xiangming |
| author_facet | Wang, Lingyu Meng, Xiangming |
| contents | Solving inverse problems with diffusion models has shown promise in tasks such as image restoration. A common approach is to formulate the problem in a Bayesian framework and sample from the posterior by combining the prior score with the likelihood score. Since the likelihood term is often intractable, estimators like DPS, DMPS, and $π$GDM are widely adopted. However, these methods rely on a fixed, manually tuned scale to balance prior and likelihood contributions. Such a static design is suboptimal, as the ideal balance varies across timesteps and tasks, limiting performance and generalization. To address this issue, we propose SAIP, a plug-and-play module that adaptively refines the scale at each timestep without retraining or altering the diffusion backbone. SAIP integrates seamlessly into existing samplers and consistently improves reconstruction quality across diverse image restoration tasks, including challenging scenarios. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_24580 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | SAIP: A Plug-and-Play Scale-adaptive Module in Diffusion-based Inverse Problems Wang, Lingyu Meng, Xiangming Machine Learning Computer Vision and Pattern Recognition Solving inverse problems with diffusion models has shown promise in tasks such as image restoration. A common approach is to formulate the problem in a Bayesian framework and sample from the posterior by combining the prior score with the likelihood score. Since the likelihood term is often intractable, estimators like DPS, DMPS, and $π$GDM are widely adopted. However, these methods rely on a fixed, manually tuned scale to balance prior and likelihood contributions. Such a static design is suboptimal, as the ideal balance varies across timesteps and tasks, limiting performance and generalization. To address this issue, we propose SAIP, a plug-and-play module that adaptively refines the scale at each timestep without retraining or altering the diffusion backbone. SAIP integrates seamlessly into existing samplers and consistently improves reconstruction quality across diverse image restoration tasks, including challenging scenarios. |
| title | SAIP: A Plug-and-Play Scale-adaptive Module in Diffusion-based Inverse Problems |
| topic | Machine Learning Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2509.24580 |