Virtual Texturing in the Documentation of Cultural Heritage

Irmengard Mayer, Claus Scheiblauer, Albert Julian Mayer
Virtual Texturing in the Documentation of Cultural Heritage
Geoinformatics FCE CTU, 7(), September 2011. [ paper draft]
Content:

Information

Abstract

In the last decade the documentation of cultural heritage by means of laser range scanning and photogrammetric techniques has gained ever more importance. The amount of data collected by these means may be huge, and adequate presentation of 3D documented cultural heritage is still a challenge. For small and limited projects consisting of only a few range scans, the software provided with the laser scanner can be used for viewing and presenting the data. Large projects, consisting of hundreds of scan positions as well as projects where more and more data are collected over time, still have to deal with a massive reduction of the 3D data for presentation. Public demonstrations in museums, as for example shown by the Digital Michelangelo project, are already state of the art. The combination of huge point-base models and mesh models with high resolution textures in one viewer, the first type of models resulting from the data of laser range scans and the second type of models resulting from a photogrammetric reconstruction process, is still not available. Currently viewers are mostly limited to show models that are based on only one geometric primitive – either points or polygons – at once. In the FWF funded START project “The Domitilla Catacomb in Rome. Archaeology, Architecture and Art History of a Late Roman Cemetery” – which is running for 5 years now – 3D point data was collected for the geometrical documentation of the vast gallery system of the Domitilla Catacomb, resulting in point data of some 2 billion (10^9) point samples. Furthermore high quality textured mesh models of the nearly 90 late Roman / early Christian paintings were generated with photogrammetric tools. In close cooperation with the Institute of Computer Graphics and Algorithms of the Vienna University of Technology the point cloud viewer Scanopy was improved for the combined presentation of huge point clouds and high quality textured mesh models in the same viewer. Our viewer is already capable of rendering huge point clouds, so for this a method to manage the vast amount of textures had to be found. Therefore we integrated a virtual texturing algorithm, which allows using the original photographs of the paintings taken on site to be mapped to the mesh models, resulting in a high quality texture for all mesh models. The photographs have a resolution of 11 Megapixels. Due to shortcomings in the programs used in the photogrammetric processing pipeline we scaled down the photographs to a 7.3 Megapixel resolution. Currently 608 of these images are used for texturing 29 mesh models. The work on the mesh models is still ongoing, and when all mesh models will be completed, we will use some 2000 images for texturing about 90 mesh models. These virtually textured models can show the details of each painting in the Domitilla Catacomb. When used in a virtual walkthrough the paintings in the catacomb can be presented to a broad audience under best lighting conditions, even the paintings normally not accessible by the public.

Additional Files and Images

Additional files:
paper draft paper draft: The Domitilla images are not included due to copyright reasons.
slides draft slides draft: The Domitilla images are not included due to copyright reasons.

BibTeX

Download BibTeX-Entry
@article{mayer_2011_cipa,
  title =      "Virtual Texturing in the Documentation of Cultural Heritage",
  author =     "Irmengard Mayer and Claus Scheiblauer and Albert Julian
               Mayer",
  year =       "2011",
  abstract =   "In the last decade the documentation of cultural heritage by
               means of laser range scanning and photogrammetric techniques
               has gained ever more importance. The amount of data
               collected by these means may be huge, and adequate
               presentation of 3D documented cultural heritage is still a
               challenge. For small and limited projects consisting of only
               a few range scans, the software provided with the laser
               scanner can be used for viewing and presenting the data.
               Large projects, consisting of hundreds of scan positions as
               well as projects where more and more data are collected over
               time, still have to deal with a massive reduction of the 3D
               data for presentation. Public demonstrations in museums, as
               for example shown by the Digital Michelangelo project, are
               already state of the art. The combination of huge point-base
               models and mesh models with high resolution textures in one
               viewer, the first type of models resulting from the data of
               laser range scans and the second type of models resulting
               from a photogrammetric reconstruction process, is still not
               available. Currently viewers are mostly limited to show
               models that are based on only one geometric primitive
               – either points or polygons – at once. In the
               FWF funded START project “The Domitilla Catacomb in
               Rome. Archaeology, Architecture and Art History of a Late
               Roman Cemetery” – which is running for 5 years
               now – 3D point data was collected for the geometrical
               documentation of the vast gallery system of the Domitilla
               Catacomb, resulting in point data of some 2 billion (10^9)
               point samples. Furthermore high quality textured mesh models
               of the nearly 90 late Roman / early Christian paintings were
               generated with photogrammetric tools. In close cooperation
               with the Institute of Computer Graphics and Algorithms of
               the Vienna University of Technology the point cloud viewer
               Scanopy was improved for the combined presentation of huge
               point clouds and high quality textured mesh models in the
               same viewer. Our viewer is already capable of rendering huge
               point clouds, so for this a method to manage the vast amount
               of textures had to be found. Therefore we integrated a
               virtual texturing algorithm, which allows using the original
               photographs of the paintings taken on site to be mapped to
               the mesh models, resulting in a high quality texture for all
               mesh models. The photographs have a resolution of 11
               Megapixels. Due to shortcomings in the programs used in the
               photogrammetric processing pipeline we scaled down the
               photographs to a 7.3 Megapixel resolution. Currently 608 of
               these images are used for texturing 29 mesh models. The work
               on the mesh models is still ongoing, and when all mesh
               models will be completed, we will use some 2000 images for
               texturing about 90 mesh models. These virtually textured
               models can show the details of each painting in the
               Domitilla Catacomb. When used in a virtual walkthrough the
               paintings in the catacomb can be presented to a broad
               audience under best lighting conditions, even the paintings
               normally not accessible by the public.",
  pages =      "%pages_from%--%pages_to%",
  month =      sep,
  event =      "XXIIIrd International CIPA Symposium",
  issn =       "1802-2669",
  journal =    "Geoinformatics FCE CTU",
  volume =     "7",
  location =   "Czech Technical University in Prague",
  keywords =   "virtual texturing, out-of-core, streaming, point clouds",
  URL =        "http://www.cg.tuwien.ac.at/research/publications/2011/mayer_2011_cipa/",
}