|
Filip Sadlo, Simon Heinzle, Tim Weyrich, Markus Gross
In this project we develop a purely point-based content creation pipeline
for 3D objects. The acquisition stage consists of a structured light scanner
built from custom hardware. The cleaning stage is implemented as a part of
the PointShop3D system. The visualization stage includes a plug-in for web
browsers that allows to make the resulting models available in the WWW.
The scanner consists of a static video projector that projects structured
light codes into the scene, a computer-controlled turntable that rotates
the scene to different views and a static camera which acquires the light
codes inside each view. Using the calibrated setup, 3D geometry and associated
texture can be reconstructed from the captured images by triangulating
the respective camera ray with the corresponding projector ray that illuminated
the surface element. For texture reconstruction, the projector is used
as a calibrated light source. Even specular texture characteristics can
be reconstructed per surface point by utilizing the different views and
illumination directions emanating from rotation by the turntable. Due to
the typically sparse reflectance sampling, only simple BRDF's conforming
to the Phong reflectance model are fitted.
Despite the steady increase in accuracy, most available scanning techniques
cause severe scanning artifacts such as noise, outliers, holes, or ghost
geometry. To apply sophisticated modeling operations on these data sets,
substantial post-processing is usually required.
In this project, we address a variety of scanning artifacts that are created
by common optical scanners and developed a comprehensive set of user-guided
tools to process corrupted data sets. The tools were implemented within a
Pointshop3D plugin. They include an eraser tool, low-pass filters for noise
removal, a set of outlier detection methods, and various up-sampling and
hole-filling tools. These techniques can be applied early in the content
creation pipeline. Therefore, all our tools are implemented to operate
directly on the acquired point cloud.
- F. Sadlo, T. Weyrich, R. Peikert, M. Gross, A Practical Structured Light Acquisition System for Point-Based Geometry and Texture, Proceedings of the Eurographics Symposium on Point-Based Graphics '05, p. 89-98 (Stony Brook, USA, June 21-22)
[Abstract]
[BibTeX]
[PDF]
- M. Pauly, N. Mitra, J. Giesen, M. Gross, L. J. Guibas, Example-Based 3D Scan Completion, Eurographics Symposium on Geometry Processing 2005
[PDF]
- T. Weyrich, M. Pauly, R. Keiser, S. Heinzle, S. Scandella, M. Gross, Post-processing of Scanned 3D Surface Data, Proceedings of Eurographics Symposium on Point-Based Graphics 2004, pp. 85-94 (Eurographics Symposium on Point-Based Graphics 2004)
[Abstract]
[PDF]
- M. Pauly, N. J. Mitra, L. J. Guibas, Uncertainty and Variability in Point Cloud Surface Data, Eurographics Symposium on Point-Based Graphics 2004
[PDF]
- D. Cotting, T. Weyrich, M. Pauly, M. Gross, Robust Watermarking of Point-Sampled Geometry, Proceedings of International Conference on Shape Modeing and Applications 2004, pp. 233-242 (SMI04, Genova, Italy, June 7-9, 2004)
[Abstract]
[PDF] [Video]
|
inf.ethz.ch