SAIP: A Plug-and-Play Scale-adaptive Module in Diffusion-based Inverse Problems
This addresses a specific bottleneck in image restoration for researchers and practitioners using diffusion models, offering an incremental improvement over existing methods.
The paper tackles the problem of suboptimal performance in diffusion-based inverse problems due to fixed scaling between prior and likelihood scores, proposing SAIP, a plug-and-play module that adaptively refines this scale at each timestep, which consistently improves reconstruction quality across diverse image restoration tasks.
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.