Diffusion Guided Upsampling of Sparse Indoor LiDAR Point Clouds for Occlusion Aware Scene Reconstruction Systems
International Journal of Information and Communication Technology Trends · 01.06.2026 · 6 (1) · 295-306
DOI: 10.66525/ijictt98 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Information and Communication Technology Trends, online erschienen am 01.06.2026. Verfasst von Claire Morel und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
The rapid advancement of three-dimensional scanning technologies has positioned Light Detection and Ranging as a cornerstone for indoor scene reconstruction, autonomous navigation, and spatial computing. However, raw point clouds acquired from indoor environments often suffer from severe sparsity, non-uniform density, and occlusion-induced missing regions, which severely bottleneck downstream applications. To address these critical limitations, this paper proposes a novel diffusion-guided point cloud upsampling framework specifically engineered for the high-fidelity reconstruction of complex indoor scenes. By formulating the upsampling task as a conditional generative process, our methodology leverages a denoising diffusion probabilistic model that operates directly on irregular, unordered point sets. The proposed architecture employs a hierarchical feature extraction mechanism coupled with cross-attention modules to capture both local geometric nuances and global structural topologies. The forward diffusion process systematically corrupts the underlying ground-truth dense point distribution, while the reverse generative phase progressively recovers high-resolution spatial coordinates conditioned on the sparse input geometry. Extensive experimental evaluations conducted on industry-standard indoor datasets demonstrate that our approach significantly outperforms existing optimization-based and generative adversarial network-based upsampling methods across multiple quantitative metrics and qualitative visual assessments. The proposed framework excels in preserving sharp edges, generating uniform point distributions, and avoiding clustering artifacts common in previous methodologies.
Zitieren
APA
Morel, C., Moretti, L., & Matsuda, R. (2026). Diffusion Guided Upsampling of Sparse Indoor LiDAR Point Clouds for Occlusion Aware Scene Reconstruction Systems. International Journal of Information and Communication Technology Trends, 6(1), 295-306. https://doi.org/10.66525/ijictt98
Vancouver
Morel C, Moretti L, Matsuda R. Diffusion Guided Upsampling of Sparse Indoor LiDAR Point Clouds for Occlusion Aware Scene Reconstruction Systems. International Journal of Information and Communication Technology Trends. 2026;6(1):295-306. doi:10.66525/ijictt98
BibTeX
@article{Morel2026,
author = {Morel, Claire and Moretti, Luca and Matsuda, Ren},
title = {Diffusion Guided Upsampling of Sparse Indoor LiDAR Point Clouds for Occlusion Aware Scene Reconstruction Systems},
journal = {International Journal of Information and Communication Technology Trends},
year = {2026},
volume = {6},
number = {1},
pages = {295--306},
doi = {10.66525/ijictt98}
}
EUVIMED