An End-to-End Geometric Deficiencies Elimination Algorithm for 3D Meshes

摘要: The 3D mesh is an important representation of geometric data. It widely used in computer graphics and has attracted more attention vision community recently. However, the generation data, deficiencies (e.g., duplicate elements, degenerate faces, isolated vertices, self-intersection, inner faces) are unavoidable. Geometric may violate topology structure object affect use meshes. In this paper, we propose end-to-end algorithm to eliminate effectively efficiently for meshes a specific reasonable order. Specifically, elements can be first eliminated by assessing appear times vertices or faces. Then, faces removed according outer product two edges. Next, since do not any face they deleted directly. Afterward, self-intersecting detected remeshed using AABB tree. Finally, detect remove whether multiple random rays shooted from reach infinity. Experiments on ModelNet40 dataset illustrate that our method thoroughly