Guidance for Low-Level Perceptual Editing in Unconditional Diffusion Models
This work provides a method for improving the aesthetic quality of images generated by unconditional diffusion models, which is a problem for users seeking refined visual outputs.
The paper addresses the challenge of aesthetically enhancing outputs from unconditional diffusion models, showing that existing h-space patching methods fail for global, low-level transformations. They propose a new training-free framework that uses degradation concept vectors and combines bottleneck patching with classifier-free guidance to steer sampling away from degraded images, resulting in consistently improved images.
Unconditional diffusion models offer powerful generative priors, yet steering them toward aesthetically enhanced outputs remains largely unexplored. We show that h-space patching, the dominant paradigm for training-free diffusion editing, systematically fails for global, low-level transformations required for aesthetic and perceptual refinement. We introduce a novel, generalized framework for image-editing in unconditional diffusion models without explicit training. This inference-time mechanism operates on low-level features by extracting degradation concept vectors and combining bottleneck patching with classifier-free guidance to guide sampling away from the degraded manifold, producing consistently improved images without any model retraining.