Computer Graphics

University of California - Berkeley

Spectral Watertight Surface Reconstruction


We use spectral partitioning to reconstruct a watertight surface from point cloud data. This method is particularly effective for noisy and undersampled point sets with outliers, because decisions about the reconstructed surface are based on a global view of the model.


Ravi Krishna Kolluri, Jonathan R. Shewchuk, and James F. O'Brien. "Spectral Watertight Surface Reconstruction". In Proceedings of SIGGRAPH 2003, Technical Sketch, July 2003.

Supplemental Material

Full Paper

The full version of this paper was published in SGP 2004.