DREAM: Diffusion Rectification and Estimation-Adaptive Models
This addresses efficiency and quality trade-offs in diffusion-based image super-resolution, though it appears incremental as an enhancement to existing methods.
The paper tackles the misalignment between training and sampling in diffusion models by introducing DREAM, a framework that requires minimal code changes and improves image super-resolution with 2-3× faster training convergence and 10-20× fewer sampling steps.
We present DREAM, a novel training framework representing Diffusion Rectification and Estimation Adaptive Models, requiring minimal code changes (just three lines) yet significantly enhancing the alignment of training with sampling in diffusion models. DREAM features two components: diffusion rectification, which adjusts training to reflect the sampling process, and estimation adaptation, which balances perception against distortion. When applied to image super-resolution (SR), DREAM adeptly navigates the tradeoff between minimizing distortion and preserving high image quality. Experiments demonstrate DREAM's superiority over standard diffusion-based SR methods, showing a $2$ to $3\times $ faster training convergence and a $10$ to $20\times$ reduction in sampling steps to achieve comparable results. We hope DREAM will inspire a rethinking of diffusion model training paradigms.