Vision, Modeling, and Visualization Workshop 2009 in Braunschweig, Germany

Title Image - "Making of" 3D-Gauss

The VMV 2009 title image shows a composite of a real world scene - captured at the Gaußhügel ('Gauss Hill') and a 3D reconstruction of the actual life-size Gauss statue.

You may download the the higher resolution versions of the title image: Textured Gauss - Wireframe Gauss


  • Acquisition

    62 photograps from the Gauss statue - located close to the university campus - were taken with a Cannon EOS 5D. The positions of the camera describe a circle around the statue.

    View Larger Map

  • Calibration

    Images are automaticaly calibrated using SIFT features [1] and the 'bundler' structure-from-motion algorithm [2]. The output data of this step contains the cameras' position, rotation and instrinsic parameters as well as a sparse set of 3D features.

  • Reconstruction

    A dense surface patch model is obtained with the Patch-Based Multi-View Stereo approach [3] . Poisson Reconstruction [4] turns the set of patches into an actual watertight mesh.

  • Compositing

    The mesh is simplified, retextured an rendered with Blender. The background is manually reconstructed in gimp using neighboring views, and the rendered statue is pasted on top of the pedestal



    by Christian Lipski





Bibliography

[1]David G. Lowe: Distinctive Image Features from Scale-Invariant Keypoints. International Journal of Computer Vision, 60, 2 (2004), pp. 91-110
[2]Noah Snavely, Steven M. Seitz, Richard Szeliski: Photo Tourism: Exploring image collections in 3D. ACM Transactions on Graphics (Proceedings of SIGGRAPH 2006)
[3]Yasutaka Furukawa and Jean Ponce: Accurate, Dense, and Robust Multi-View Stereopsis . CVPR 2007
[4]M. Kazhdan, M. Bolitho, and H. Hoppe: Poisson Surface Reconstruction . Symposium on Geometry Processing (June 2006), pages 61--70