Point Cloud Outlier Removal with Corruption-Stable Tokenization
Electronics · 2026-11-13 · 15 (22) · 5272
DOI: 10.3390/electronics15225272 ↗At a glance
Journal article in Electronics, published online on 2026-11-13. Written by Maoz Koren and 2 co-authors. 21 references. The full text is freely available (open access, CC BY 4.0).
Abstract (original language)
Real-world point clouds often contain spurious observations that obscure the underlying geometry. Modern point-cloud models struggle to distinguish such outliers from sparse but valid local geometry. We propose a two-stage transformer framework for additive outlier removal that combines self-supervised masked reconstruction with supervised point-wise classification. To reduce pre-encoding corruption, the framework replaces data-dependent token centers with fixed spatial centers, drawing on Vision Transformers. This establishes a stable token layout while still allowing token contents to vary. The model retains points classified as structural without changing their coordinates. Experiments on normalized object-level point clouds with synthetic additive corruption show that the method removes a substantial proportion of outliers while preserving most structural points. Self-supervised pretraining is critical to this balance, while alternative tokenization changes the removal-preservation trade-off. Performance remains stable across tested contamination levels, although near-surface outliers and thin or sparsely sampled structures remain challenging.
Cite
APA
Koren, M., Yellin, D., & Aperstein, Y. (2026). Point Cloud Outlier Removal with Corruption-Stable Tokenization. Electronics, 15(22), 5272. https://doi.org/10.3390/electronics15225272
Vancouver
Koren M, Yellin D, Aperstein Y. Point Cloud Outlier Removal with Corruption-Stable Tokenization. Electronics. 2026;15(22):5272. doi:10.3390/electronics15225272
BibTeX
@article{Koren2026,
author = {Koren, Maoz and Yellin, Dovi and Aperstein, Yehudit},
title = {Point Cloud Outlier Removal with Corruption-Stable Tokenization},
journal = {Electronics},
year = {2026},
volume = {15},
number = {22},
pages = {5272},
doi = {10.3390/electronics15225272}
}
EUVIMED