CAREER: Signal Processing Tools for Dynamic Geometry

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

0133554 Guskov, Igor U of MIchigan-Ann Arbor Medical imaging systems and computational modeling of physical processes produce vast amounts of surface data for analysis and visualization purposes. Often such data are time-dependent and describe surfaces that evolve in time, for example, a developing interface between two fluids or moving walls of the heart ventricles. This project will develop processing tools for such dynamic surface data. In particular, massively redundant and noisy raw data coming from applications will be converted into compact form useful for extensive interactive manipulation, especially within remote collaboration environments. The developed technology will also be useful for applications in tele-presence and animation. This research effort will be integrated with educational initiatives via student involvement in research projects, development of new curriculum in Digital Geometry Processing, and collaboration with external researchers. The focus of this project will be on converting raw application data into hierarchical surface representations that are temporally coherent and spatially (semi-)regular, and further processing and compression of the constructed data representations. Specifically, a sequence of irregular meshes produced by a frame-by-frame marching cubes extraction of isosurfaces from a time-dependent volumetric field is an example of geometric data that cannot be efficiently processed without first establishing temporal coherence and spatial regularity of the surface sampling. The development of a fully automated compression system for dynamic geometry in this scenario will include the following components: constrained parameterization method for establishing temporal coherence of surface sampling, patch layout generation and maintenance routine to deal with possible topology changes, efficient topological event detection procedure, predictive coding of dynamic geometry streams, geometric and topological denoising of raw application data, and possible integration of the regular mesh extraction into the simulation (or shape acquisition) environment. The developed technology will also be useful in the areas of scientific visualization and computer aided geometric design.