Generalized unstructured decimation [computer graphics]

The research focuses on vertex removal methodology, not on the formulation of decimation criteria, and the resulting decimation algorithm is robust and efficient because it avoids floating-point classification computations.

Abstract

Decimation describes the process of removing entities (such as polygons) from a geometric representation. The goal is to intelligently reduce the number of primitives required to accurately model the problem of interest. The work described in the article was originally motivated by the need for efficient and robust decimation of volume tessellations, that is, unstructured tetrahedrizations. Existing surface-based decimation schemes do not generalize to volumes. The technique allows local, dynamic vertex removal from an unstructured tetrahedrization while preserving the initial tessellation topology and boundary geometry. The research focuses on vertex removal methodology, not on the formulation of decimation criteria. In practice, criteria for removing vertices are application specific. The basis of the algorithm is a unique and general method to classify a triangle with respect to a nonconvex polygon. The resulting decimation algorithm (applicable to both surface and volume tessellations) is robust and efficient because it avoids floating-point classification computations.