HP-PnP-Flow: Hybrid Time Scheduling and Projection for Plug-and-Play Image Restoration via Flow Matching

Key points

• Image restoration is a fundamental inverse problem that seeks to recover high-quality images from degraded observations.1
• Numerical experiments on CelebA and AFHQ-Cat demonstrate that the newly proposed HP-PnP-Flow algorithm significantly outperforms PnP-Flow on several restoration tasks such as denoising, deblurring, super-resolution, random inpainting, and box inpainting.1
• Notably, our algorithm achieves comparable or better reconstruction quality with substantially fewer iterations, reducing the iteration count from $500$ to $100$ for deblurring and super-resolution on AFHQ-Cat.1

Sources (1)

• [1]HP-PnP-Flow: Hybrid Time Scheduling and Projection for Plug-and-Play Image Restoration via Flow Matching
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 11:15 AM

  Plug-and-Play (PnP) methods provide a flexible framework for integrating pretrained denoisers into iterative reconstruction, whereas flow matching offers an efficient generative modeling approach based on continuous-time ODE dynamics.

  Image restoration is a fundamental inverse problem that seeks to recover high-quality images from degraded observations.

Extractive summary: sentences quoted from the sources.