Faithful Contouring: Near-Lossless 3D Voxel Representation Free from Iso-surface
This work addresses the need for high-fidelity 3D representations in reconstruction and generation tasks, offering a novel approach that avoids common limitations like water-tightening, but it is incremental as it builds on existing voxel-based methods with specific improvements.
The paper tackles the problem of accurate and efficient voxelized representations for 3D meshes, which often compromise geometric fidelity in existing methods, by proposing Faithful Contouring, a sparse voxelized representation that achieves near-lossless fidelity with distance errors at the 10^-5 level and reduces Chamfer Distance by 93% and improves F-score by 35% over baselines for mesh reconstruction.
Accurate and efficient voxelized representations of 3D meshes are the foundation of 3D reconstruction and generation. However, existing representations based on iso-surface heavily rely on water-tightening or rendering optimization, which inevitably compromise geometric fidelity. We propose Faithful Contouring, a sparse voxelized representation that supports 2048+ resolutions for arbitrary meshes, requiring neither converting meshes to field functions nor extracting the isosurface during remeshing. It achieves near-lossless fidelity by preserving sharpness and internal structures, even for challenging cases with complex geometry and topology. The proposed method also shows flexibility for texturing, manipulation, and editing. Beyond representation, we design a dual-mode autoencoder for Faithful Contouring, enabling scalable and detail-preserving shape reconstruction. Extensive experiments show that Faithful Contouring surpasses existing methods in accuracy and efficiency for both representation and reconstruction. For direct representation, it achieves distance errors at the $10^{-5}$ level; for mesh reconstruction, it yields a 93\% reduction in Chamfer Distance and a 35\% improvement in F-score over strong baselines, confirming superior fidelity as a representation for 3D learning tasks.