3D Shape Modeling with Adaptive Centroidal Voronoi Tesselation on Signed Distance Field
Résumé
Volumetric shape representations have become ubiquitous in multi-view reconstruction tasks. They often build on regular voxel grids as discrete representations of 3D shape functions, such as SDF or radiance fields, either as the full shape model or as sampled instantiations of continuous representations, as with neural networks. Despite their proven efficiency, voxel representations come with the precision versus complexity tradeoff. This inherent limitation can significantly impact performance when moving away from simple and uncluttered scenes. In this paper we investigate an alternative discretization strategy with the Centroidal Voronoi Tesselation (CVT). CVTs allow to better partition the observation space with respect to shape occupancy and to focus the discretization around shape surfaces. To leverage this discretization strategy for multi-view reconstruction, we introduce a volumetric optimization framework that combines explicit SDF fields with shallow color networks, in order to estimate 3D shape properties over tetrahedral grids. Experiments validate this approach with significantly better reconstruction quality on large scenes while remaining competitive with small scenes. In addition to reconstruction improvements, CVTs also yield a higher quality alternative to marching cubes when meshing volumetric shape representations.
Origine | Fichiers produits par l'(les) auteur(s) |
---|