PhysGasFluid: Physics-Guided Gaseous Fluid Flow Reconstruction

I. Overview

Gaseous fluids are semi-transparent, deform continuously, and often provide only weak visual cues. PhysGasFluid studies how physical knowledge can guide the reconstruction of this evolving flow instead of relying on appearance alone.

II. Research Focus

The project focuses on recovering plume geometry and motion while preserving fine, low-contrast structures. Physics-guided learning provides constraints that complement the incomplete evidence available in captured gas imagery.

III. Method Overview

PhysGasFluid combines data-driven reconstruction with physical guidance for gaseous flow. The supplied overview compares its recovered plumes with FluidNexus and ground truth at several moments, including magnified regions that expose differences in local structure.

IV. Publication

This work is scheduled to appear at the 2026 IEEE International Conference on Image Processing.

@inproceedings{wu2026physgasfluid,
  title={PhysGasFluid: Physics-Guided Gaseous Fluid Flow Reconstruction},
  author={Wu, Keyi and Du, Shan},
  booktitle={2026 IEEE International Conference on Image Processing (ICIP)},
  pages={1--6},
  year={2026},
  organization={IEEE}
}