Robust Efficient Medial Axis Computation of Loop Subdivision Surfaces

Speaker: Leo Preissegger

Abstract

We propose a robust and efficient method for computing the medial axis of Loop subdivision limit surfaces of a closed triangle control mesh. By evaluating the exact limit surface, including at regions around extraordinary vertices, and using a branch-and-bound search for distance queries to the surface, our method avoids the artifacts and incorrect closest points of previous approximation-based approaches. The same distance queries also allow the limit surface to be evaluated as a SDF. Through GPU acceleration, this technique makes dense medial axis sampling practical and enables accurate local feature size computation for applications such as local feature size aware sampling and remeshing.