Share Email Print

Proceedings Paper

3D polygonal representation of dense point clouds by triangulation, segmentation, and texture projection
Author(s): Touraj Tajbakhsh
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

A basic concern of computer graphic is the modeling and realistic representation of three-dimensional objects. In this paper we present our reconstruction framework which determines a polygonal surface from a set of dense points such those typically obtained from laser scanners. We deploy the concept of adaptive blobs to achieve a first volumetric representation of the object. In the next step we estimate a coarse surface using the marching cubes method. We propose to deploy a depth-first search segmentation algorithm traversing a graph representation of the obtained polygonal mesh in order to identify all connected components. A so called supervised triangulation maps the coarse surfaces onto the dense point cloud. We optimize the mesh topology using edge exchange operations. For photo-realistic visualization of objects we finally synthesize optimal low-loss textures from available scene captures of different projections. We evaluate our framework on artificial data as well as real sensed data.

Paper Details

Date Published: 4 February 2010
PDF: 8 pages
Proc. SPIE 7526, Three-Dimensional Image Processing (3DIP) and Applications, 75260C (4 February 2010); doi: 10.1117/12.840980
Show Author Affiliations
Touraj Tajbakhsh, Silicon Image GmbH (Germany)

Published in SPIE Proceedings Vol. 7526:
Three-Dimensional Image Processing (3DIP) and Applications
Atilla M. Baskurt, Editor(s)

© SPIE. Terms of Use
Back to Top