Robust Voxelization and Visualization by Improved Tetrahedral Mesh Generation

TL;DR

This paper introduces a robust voxelization framework that converts triangular meshes into voxel data using high-quality tetrahedral mesh generation, outperforming existing methods especially on low-quality meshes.

Contribution

It presents a novel tetrahedral mesh generation method that improves voxelization quality and robustness, with GPU and CPU parallelization, validated on diverse datasets.

Findings

Near 100% voxelization success on tested datasets

Outperforms state-of-the-art methods in quality and robustness

Effective on low-quality meshes from internet sources

Abstract

When obtaining interior 3D voxel data from triangular meshes, most existing methods fail to handle low quality meshes which happens to take up a big portion on the internet. In this work we present a robust voxelization method that is based on tetrahedral mesh generation within a user defined error bound. Comparing to other tetrahedral mesh generation methods, our method produces much higher quality tetrahedral meshes as the intermediate outcome, which allows us to utilize a faster voxelization algorithm that is based on a stronger assumption. We show the results comparing to various methods including the state-of-the-art. Our contribution includes a framework which takes triangular mesh as an input and produces voxelized data, a proof to an unproved algorithm that performs better than the state-of-the-art, and various experiments including parallelization built on the GPU and CPU. We further tested our method on various dataset including Princeton ModelNet and Thingi10k to show the robustness of the framework, where near 100% availability is achieved, while others can only achieve around 50%.