Manifold-guaranteed out-of-core simplification of large meshes with controlled topological type

Abstract

In this paper, a simple and efficient algorithm is proposed for manifold-guaranteed out-of-core simplification of large meshes with controlled topological type. By dual-sampling the input large mesh model, the proposed algorithm utilizes a set of Hermite data (sample points with normals) as an intermediate model representation, which allows the topological structure of the mesh model to be encoded implicitly and thus makes it particularly suitable for out-of-core mesh simplification. Benefiting from the construction of an in-core signed distance field, the proposed algorithm possesses a set of features including manifoldness of the simplified meshes, toleration of nonmanifold mesh data input, topological noise removal, topological type control and, sharp features and boundary preservation. A novel, detailed implementation of the proposed algorithm is presented, and experimental results demonstrate that very large meshes can be simplified quickly on a low-cost off-the-shelf PC with tightly bounded approximation errors and with time and space efficiency.