Manifold Contouring of an Adaptively Sampled Distance Field

Abstract

Volumetric data is a convenient representation of shape on many occasions. One application area is remeshing, where a poorly triangulated model is converted to a volumetric representation and then transformed back into a model of better triangle quality. In certain areas, for example medical scans, volumetric data arise naturally. To render the shapes captured by such a scan, a common approach is to convert the volumetric data into a triangle mesh. Since both types of representations are valuable, it is interesting to find reliable and effcient ways of converting between them. Here, we will exclusively look at the conversion from volume data to triangle mesh. Many methods exist for performing such an operation, where one of the most popular is Marching Cubes. The meshes resulting from this algorithm will however have properties often undesirable. Furthermore, the method is not applicable when storing the volumetric data in an adaptive structure, such as an octree. In this report, we will describe how meshes without these undesirable properties can be generated from an octree in a straightforward manner.