Weighted mesh simplification of 3D triangular surfaces

Abstract

Representing huge triangular datasets in a real-time rendering environment is a challenge which receives continuous attention. There is a growing complexity of geometric meshes on the one hand and an increasing computational power of graphics hardware on the other hand.<br />Although hardware acceleration is very powerful, simplification using software is much cheaper and more flexible. Additionally, for a large class of geometric models, simplification can be performed as a preprocessing step where there is no demand for real-time operations.<br />The underlying theory of this diploma thesis is based on mesh simplification using quadric error metrics. This algorithm was first published by Michael Garland and it was implemented into Aardvark, a sophisticated rendering framework. It contains a comprehensive set of libraries dealing with data structures in general and polygonal mesh data structures in particular.<br />The application takes a triangulated mesh as input and iteratively creates an even more simplified approximation by weighting and collapsing suitable mesh areas. Some surface details will be lost, but the overall shape of the model will be preserved. The presented code can handle models by preprocessing static, closed triangular meshes. The iterative computation stops when a user specified percentage of the original mesh size is reached.<br />A further improvement in a later research project will try to add vertex colors, normals and texture coordinates to the simplification method.