Physics-Consistent Diffusion for Efficient Fluid Super-Resolution via Multiscale Residual Correction

Researchers developed ReMD, a physics-consistent diffusion framework that improves fluid super-resolution by incorporating physical constraints and multiscale modeling. The approach addresses limitations of existing image and diffusion models when applied to fluid dynamics, achieving better accuracy and spectral fidelity with fewer sampling steps.

• →ReMD introduces a novel physics-consistent diffusion framework specifically designed for fluid super-resolution tasks.
• →The method uses multigrid residual correction and multi-wavelet basis to capture both large structures and fine details in fluid dynamics.
• →ReMD achieves comparable quality to existing diffusion models while requiring significantly fewer sampling steps.
• →The framework addresses common issues in fluid simulation including spectral mismatch and spurious divergence.
• →Testing on atmospheric and oceanic benchmarks demonstrates improved accuracy and spectral fidelity compared to baseline models.