Interactive Screen-Space Triangulation for High-Quality Rendering of Point Clouds

Information

• Publication Type: Technical Report
• Workgroup(s)/Project(s):
  • TERAPOINTS - High Quality Visualization and Interaction with Gigantic Point Data Sets
• Date: April 2012
• Number: TR-186-2-12-01
• Keywords: screen triangulation, point rendering, nearest neighbors, screen-space, point clouds

Abstract

This technical report documents work that is a precursor to the Auto Splatting technique. We present a rendering method that reconstructs high quality images from unorganized colored point data. While previous real-time image reconstruction approaches for point clouds make use of preprocessed data like point radii or normal estimations, our algorithm only requires position and color data as input and produces a reconstructed color image, normal map and depth map which can instantly be used to apply further deferred lighting passes. Our method performs a world-space neighbor search and a subsequent normal estimation in screen-space, and uses the geometry shader to triangulate the color, normal and depth information of the points. To achieve correct visibility and closed surfaces in the projected image a temporal coherence approach reuses triangulated depth information and provides adaptive neighbor search radii. Our algorithm is especially suitable for insitu high-quality visualization of big datasets like 3D-scans, making otherwise time-consuming preprocessing steps to reconstruct surface normals or point radii dispensable.

Additional Files and Images

Weblinks

No further information available.

BibTeX

@techreport{TR-186-2-12-01,
  title =      "Interactive Screen-Space Triangulation for High-Quality
               Rendering of Point Clouds",
  author =     "Reinhold Preiner and Michael Wimmer",
  year =       "2012",
  abstract =   "This technical report documents work that is a precursor to
               the Auto Splatting technique. We present a rendering method
               that reconstructs high quality images from unorganized
               colored point data. While previous real-time image
               reconstruction approaches for point clouds make use of
               preprocessed data like point radii or normal estimations,
               our algorithm only requires position and color data as input
               and produces a reconstructed color image, normal map and
               depth map which can instantly be used to apply further
               deferred lighting passes. Our method performs a world-space
               neighbor search and a subsequent normal estimation in
               screen-space, and uses the geometry shader to triangulate
               the color, normal and depth information of the points. To
               achieve correct visibility and closed surfaces in the
               projected image a temporal coherence approach reuses
               triangulated depth information and provides adaptive
               neighbor search radii. Our algorithm is especially suitable
               for insitu high-quality visualization of big datasets like
               3D-scans, making otherwise time-consuming preprocessing
               steps to reconstruct surface normals or point radii
               dispensable.",
  month =      apr,
  number =     "TR-186-2-12-01",
  address =    "Favoritenstrasse 9-11/E193-02, A-1040 Vienna, Austria",
  institution = "Institute of Computer Graphics and Algorithms, Vienna
               University of Technology ",
  note =       "human contact: technical-report@cg.tuwien.ac.at",
  keywords =   "screen triangulation, point rendering, nearest neighbors,
               screen-space, point clouds",
  URL =        "https://www.cg.tuwien.ac.at/research/publications/2012/TR-186-2-12-01/",
}