ESSIC | State Climatologist | Fellowships | Ocean Sciences at UMD | AOSC Grads |
AOSC Departmental Seminar
February 18, 2016

Representation and topological analysis for scientific data visualization

Leila De Floriani
University of Maryland Institute for Advanced Computer Studies, University of Maryland

The visualization of currently available scientific datasets (describing terrains, 3D scalar or vector fields, etc.), usually of very large size,  requires effective spatial  data structures for efficiently encoding such data sets  as well as powerful and compact topological descriptors.  Data representation and topology-based visualization are two fundamental lines of the current research in scientific data visualization.  The first part of the talk will focus  on two recent contributions of our research on efficient  spatial data structures for representing scalar fields: Sparse Pyramids, compact multi-resolution  representations for terrains   and volume datasets, and  the Stellar Tree, a compact  spatio-topological  data structure for irregularly distributed data, naturally geared to a parallel and distributed computation. The second part of the talk will discuss  topology-based techniques for data visualization, and, specifically,  morphological segmentations based on discrete Morse theory, called Morse-Smale complexes. A discrete  gradient field is computed from the scalar field values,  and simplified in order to eliminate critical points  that arise on account of noise, or that correspond to uninteresting morphological features. Algorithms for building and simplifying a discrete Morse gradient, and for computing a  Morse-Smale complex are presented. Applications of such topological descriptors will also be discussed.

(This is joint work with Federico Iuricich, Riccardo Fellegara, Paola Magillo, Kenneth Weiss)