AddSR: Accelerating Diffusion-based Blind Super-Resolution with Adversarial Diffusion Distillation
This work addresses the computational bottleneck for researchers and practitioners in image processing, enabling faster high-quality image upscaling, though it is incremental as it builds on existing distillation and ControlNet techniques.
The paper tackles the poor efficiency of diffusion-based blind super-resolution methods, which require many sampling steps, by proposing AddSR, a method that accelerates the process while maintaining or improving image quality, achieving up to 7x faster speed than previous state-of-the-art models.
Blind super-resolution methods based on stable diffusion showcase formidable generative capabilities in reconstructing clear high-resolution images with intricate details from low-resolution inputs. However, their practical applicability is often hampered by poor efficiency, stemming from the requirement of thousands or hundreds of sampling steps. Inspired by the efficient adversarial diffusion distillation (ADD), we design~\name~to address this issue by incorporating the ideas of both distillation and ControlNet. Specifically, we first propose a prediction-based self-refinement strategy to provide high-frequency information in the student model output with marginal additional time cost. Furthermore, we refine the training process by employing HR images, rather than LR images, to regulate the teacher model, providing a more robust constraint for distillation. Second, we introduce a timestep-adaptive ADD to address the perception-distortion imbalance problem introduced by original ADD. Extensive experiments demonstrate our~\name~generates better restoration results, while achieving faster speed than previous SD-based state-of-the-art models (e.g., $7$$\times$ faster than SeeSR).