Information

  • Publication Type: Journal Paper with Conference Talk
  • Workgroup(s)/Project(s):
  • Date: November 2020
  • Journal: Computer Graphics Forum
  • Volume: 39
  • Open Access: yes
  • Number: 7
  • Location: Wellington, NZ
  • Lecturer: Markus Schütz
  • ISSN: 1467-8659
  • Event: Pacific Graphics 2020
  • DOI: 10.1111/cgf.14134
  • Call for Papers: Call for Paper
  • Pages: 13
  • Publisher: John Wiley & Sons, Inc.
  • Conference date: 2021
  • Pages: 1 – 13
  • Keywords: point clouds, point-based rendering, level of detail

Abstract

We propose an efficient out-of-core octree generation method for arbitrarily large point clouds. It utilizes a hierarchical counting sort to quickly split the point cloud into small chunks, which are then processed in parallel. Levels of detail are generated by subsampling the full data set bottom up using one of multiple exchangeable sampling strategies. We introduce a fast hierarchical approximate blue-noise strategy and compare it to a uniform random sampling strategy. The throughput, including out-of-core access to disk, generating the octree, and writing the final result to disk, is about an order of magnitude faster than the state of the art, and reaches up to around 6 million points per second for the blue-noise approach and up to around 9 million points per second for the uniform random approach on modern SSDs.

Additional Files and Images

Additional images and videos


Additional files

Weblinks

BibTeX

@article{SCHUETZ-2020-MPC,
  title =      "Fast Out-of-Core Octree Generation for Massive Point Clouds",
  author =     "Markus Sch\"{u}tz and Stefan Ohrhallinger and Michael Wimmer",
  year =       "2020",
  abstract =   "We propose an efficient out-of-core octree generation method
               for arbitrarily large point clouds. It utilizes a
               hierarchical counting sort to quickly split the point cloud
               into small chunks, which are then processed in parallel.
               Levels of detail are generated by subsampling the full data
               set bottom up using one of multiple exchangeable sampling
               strategies. We introduce a fast hierarchical approximate
               blue-noise strategy and compare it to a uniform random
               sampling strategy. The throughput, including out-of-core
               access to disk, generating the octree, and writing the final
               result to disk, is about an order of magnitude faster than
               the state of the art, and reaches up to around 6 million
               points per second for the blue-noise approach and up to
               around 9 million points per second for the uniform random
               approach on modern SSDs.",
  month =      nov,
  journal =    "Computer Graphics Forum",
  volume =     "39",
  number =     "7",
  issn =       "1467-8659",
  doi =        "10.1111/cgf.14134",
  pages =      "13",
  publisher =  "John Wiley & Sons, Inc.",
  pages =      "1--13",
  keywords =   "point clouds, point-based rendering, level of detail",
  URL =        "https://www.cg.tuwien.ac.at/research/publications/2020/SCHUETZ-2020-MPC/",
}