We describe two current computer graphics projects at UT Austin.

The first one is a new method for reconstructing two-dimensional surfaces from scattered data points in 3D. This is a fundamental computational problem in building computer models from captured data.

We have a simple Voronoi-based algorithm that is provably correct in the following sense: given a dense enough sample from a smooth surface, we output a set of triangles that forms a surface (which we call the "crust") homeomorphic to the original smooth surface. Our definition of "dense enough" requires lots of samples around small details but might be sparse elsewhere. The new algorithm is easy to implement and easy to use.

The second project is an image-based rendering system which produces real-time views of arbitrarily complicated scenes from a collection of stored images. The images are chosen so as to uniformly sample the space of all possible lines incident to an object, allowing smooth arbitrary camera motions. Other features of the system include: depth correction, hierarchical multiresolution, and smooth transitions between levels of detail.

The reconstruction work is due to Nina Amenta, Marshall Bern (Xerox PARC), David Eppstein (UC Irvine) and Manolis Kamvysellis (MIT). The image-based rendering work is due to Emilio Camahort and Don Fussell.