Computational fluid dynamics (CFD) now generates terabytes to
petabytes of data, creating the daunting task of trying to find, extract, and analyze important flow features (e.g., time varying vortices, shock waves) buried within these monstrous datasets.
Our project focused on developing new techniques for enabling analysis, insight, and hypothesis testing from these massive computational fluid dynamics datasets. We researched the approach of procedural encoding and real-time reconstruction to help scientists analyze and understand these large datasets.
The techniques we developed led to broader impacts than CFD, especially when used to represent scalar field isosurfaces evolution, or when applied to the optical flow of video. This supported results in a new procedural version of the level set method and to the patented TextureShop and RotoTexture methods for texturing surfaces in photos and videos.