Fast Image Super-Resolution via Consistency Rectified Flow
Researchers have developed FlowSR, a new method for image super-resolution that significantly speeds up the process using diffusion models. By reformulating super-resolution as a rectified flow from low-resolution to high-resolution images, FlowSR achieves high-quality results in a single step. The approach incorporates an improved consistency learning strategy with HR regularization and a fast-slow scheduling technique to balance efficiency and detail. AI
IMPACT This new method could enable faster and more efficient image upscaling in various applications.