Efficient Representation of Massive Geometric Models

University Of Illinois At Urbana-Champaign, Urbana IL

Investigators

Abstract

Proposal #0098170 U of Ill Urbana-Champaign Michael Garland Numerous graphics applications in areas ranging from CAD/CAM to realistic immersive simulators rely on increasingly complex datasets to achieve convincing levels of visual realism. However, the enormity of the raw geometric data frequently makes it impossible to efficiently process such datasets given limited hardware capacity. Surface models containing millions of triangles are now commonplace, and advances in acquisition technology are making models containing several billion triangles available. Consequently, there has been considerable interest over the last decade in techniques for the automatic simplification of highly detailed polygonal models. However, current methods are, almost without exception, completely incapable of processing input models of this enormous magnitude. This is a very serious shortcoming, as these are exactly the class of models for which effective simplification methods are most pressingly needed. The goal of this project is to develop new techniques for representing and processing very large scale polygonal surface models, enabling the efficient use of extremely complex models far beyond the capability of current systems. Algorithmic scalability is essential in this domain. This research is focused on developing simplification methods which combine simple out-of-core data operations with more complex output-sensitive (i.e., dependent only on the output, rather than the input, size) processing phases. The general approach of this project is to adopt recursive partitioning strategies directed by quadric error metrics. An approximation can be produced from any partition of the vertex set by merging all vertices within each cell of the partition. The use of quadric error metrics means that the aggressive simplification methods designed for this project can be seamlessly coupled with other quadric-based simplification algorithms in a multi-phase process.