Contextual Chamfer Distance for Point Cloud Restoration
IEEE Transactions on Pattern Analysis and Machine Intelligence · 2026 · 1-18
DOI: 10.1109/tpami.2026.3752196 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Pattern Analysis and Machine Intelligence, 2026. Verfasst von Zhiheng Fu und 8 weiteren Autor:innen.
Abstract (Originalsprache)
Accurate similarity measurement between point clouds is crucial for 3D geometry restoration. Chamfer Distance (CD) and its variants quantify such similarity in a permutation-invariant manner, yet they emphasize low-order point-wise proximity, leading to clustered distributions and long-range structural distortions that undermine geometric fidelity. This work rethinks point cloud similarity from a high-order standpoint and introduces Contextual Chamfer Distance (CCD), a geometrically consistent formulation that jointly models point-wise proximity, local distribution consistency, and long-range structural consistency. Specifically, Patch Density Consistency (PDC) aligns relative patch densities between predicted and ground-truth points, promoting spatial coverage for geometric completeness, while Anchor Normal Consistency (ANC) models long-range relationships through anchor normals derived from non-degenerate point triplets, improving structural coherence and geometric accuracy. Alongside this formulation, we develop a conditional local coordinate-space stability and convergence analysis and derive sufficient conditions for local linear convergence of the corresponding weighted objective under explicit local assumptions. Extensive experiments on completion, compression, and upsampling across architectures and six datasets demonstrate consistent quantitative and qualitative gains across settings, a favorable accuracy-efficiency trade-off, and strong generalization.
Zitieren
APA
Fu, Z., Cai, S., Wang, B., Guo, Y., Wang, L., Xu, L., Laga, H., Boussaid, F., & Bennamoun, M. (2026). Contextual Chamfer Distance for Point Cloud Restoration. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1-18. https://doi.org/10.1109/tpami.2026.3752196
Vancouver
Fu Z, Cai S, Wang B, Guo Y, Wang L, Xu L, et al. Contextual Chamfer Distance for Point Cloud Restoration. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2026:1-18. doi:10.1109/tpami.2026.3752196
BibTeX
@article{Fu2026,
author = {Fu, Zhiheng and Cai, Sudong and Wang, Bing and Guo, Yulan and Wang, Longguang and Xu, Lian and Laga, Hamid and Boussaid, Farid and Bennamoun, Mohammed},
title = {Contextual Chamfer Distance for Point Cloud Restoration},
journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
year = {2026},
pages = {1--18},
doi = {10.1109/tpami.2026.3752196}
}
EUVIMED