Multiattribute noise in photogrammetric point clouds: A critical review of denoising frameworks and identified gaps
Multidisciplinary Reviews · 14.10.2026 · 10 (4) · 2027231
DOI: 10.31893/multirev.2027231 ↗Kurz gefasst
Fachartikel in der Zeitschrift Multidisciplinary Reviews, online erschienen am 14.10.2026. Verfasst von Yun Sin Chong und 2 weiteren Autor:innen. Die Arbeit nennt 82 Literaturangaben.
Abstract (Originalsprache)
Photogrammetric point cloud reconstruction using consumer-grade devices has become a widely adopted, low-cost alternative to active sensing, enabling applications in heritage documentation, digital twins, virtual tourism, and serious games. However, the noise inherent in these pipelines remains poorly characterized and inadequately addressed by existing denoising methods. In this review, we systematically characterize photogrammetric noise as a multiattribute phenomenon, encompassing geometric displacement, surface orientation corruption, and chromatic inconsistency. This is further exacerbated by nonuniform point distribution due to occlusions and variable image coverage. These noise properties arise from three convergent sources: sensor-level optical distortions and pixel limitations, acquisition-induced instability and illumination variability, and inconsistent image overlap leading to density nonuniformity. These characteristics are especially pronounced in consumer-grade workflows and are not adequately represented by the synthetic noise models underlying most current denoising methods. Among existing approaches, filtering-based methods are valued for their conceptual simplicity, efficiency, and local applicability, making them the most practical choice for consumer-grade deployment. A structured evaluation of filtering frameworks against the identified noise profile reveals five critical gaps: attribute blindness, normal instability, operational complexity, structural rigidity, and lack of density adaptivity. Optimization-based and deep learning approaches, although conceptually distinct, are limited by data and computational requirements, reducing their suitability for consumer-grade workflows. This study is novel in systematically evaluating multiple frameworks against a photogrammetric-specific noise profile, offering a cross-framework characterization not previously reported. Together, these findings reveal a fundamental misalignment between existing denoising logic and the multiattribute, nonuniform noise of consumer-grade photogrammetric data. They establish the criteria that future denoising frameworks must satisfy to be practically deployable in passive, image-based reconstruction pipelines.
Zitieren
APA
Chong, Y. S., Wang, H. H., & Wang, Y. C. (2026). Multiattribute noise in photogrammetric point clouds: A critical review of denoising frameworks and identified gaps. Multidisciplinary Reviews, 10(4), 2027231. https://doi.org/10.31893/multirev.2027231
Vancouver
Chong YS, Wang HH, Wang YC. Multiattribute noise in photogrammetric point clouds: A critical review of denoising frameworks and identified gaps. Multidisciplinary Reviews. 2026;10(4):2027231. doi:10.31893/multirev.2027231
BibTeX
@article{Chong2026,
author = {Chong, Yun Sin and Wang, Hui Hui and Wang, Yin Chai},
title = {Multiattribute noise in photogrammetric point clouds: A critical review of denoising frameworks and identified gaps},
journal = {Multidisciplinary Reviews},
year = {2026},
volume = {10},
number = {4},
pages = {2027231},
doi = {10.31893/multirev.2027231}
}
EUVIMED