|Date:||Tue, September 11, 2012|
|Place:||Research I Seminar Room|
Abstract: Polygonal meshes, particularly triangular and quadrilateral, are a standard data structure used to model surfaces. The piecewise linear surface described by a mesh is the union of the convex combinations of connected vertices. Subdivision is as an alternate way to interpret the connectivity data of a polygonal mesh. Subdivision uses the vertex and connectivity information to produce a finer mesh. In the limit (assuming convergence) subdivision defines a generally smoother surface than the piecewise linear surface. We can provide additional degrees of freedom to the process by allowing displacement vectors to be defined at each vertex on each level of subdivision. In particular, by restricting those vectors to have a direction which can be predicted by the coarse mesh, such a displacement vector can be described by a single value (the detail). This allows us to construct high resolution shapes using fewer parameters than the equivalent polygonal mesh. In this project a software tool was developed to predict and find the details required to construct such a high resolution shape, be it defined implicitly or by a high resolution target mesh. An overall description of the process, the technical difficulties, and some examples (of where it works and does not work) will be discussed in this talk.