A Bayesian Method for Probable Surface Reconstruction and Decimation James Diebel and Sebastian Thrun, Stanford University, Palo Alto, CA Michael Breunig, Bosch Research, Palo Alto, CA

Please see the research page for details on other projects.

Abstract— We present a Bayesian technique for the reconstruction and subsequent decimation of 3D surface models from noisy sensor data. The method uses steerable probabilistic models of the measurement noise, and combines them with feature-enhancing prior probabilities over 3D surfaces. When applied to surface reconstruction, the method simultaneously smooths noisy regions while enhancing features, such as corners. When applied to surface decimation, it finds models that closely approximate the original mesh when rendered. The method is applied in the context of computer animation, where it finds decimations that minimize the visual error even under non-rigid deformations.

This paper was published in the January 2006 edition of ACM Transactions on Graphics (TOG).

	Paper: ACM Transactions on Graphics Paper [PDF] [Bibtex].	2/1/2006
	Movie: Example of smoothing and decimation of an active stereo vision scan of a styrofoam cup.	1/5/2005
	Movie: Example of smoothing a synthetically-generated scan of a corner.	1/5/2005
	Movie: Example of smoothing and decimation of an active stereo vision scan of a robotic toy dog.	1/5/2005
	Movie: Example of decimating the Stanford bunny.	1/5/2005
	Movie: Example of smoothing an active stereo scan of a lotion bottle.	12/18/2004
	Movie: Example of the same lotion bottle smoothed, but showing both the original noisy mesh and the smoothing wireframe simultaneously.	12/18/2004
	Movie: Example of smoothing a synthetically-gnerated scan of a corner.	12/18/2004
	Movie: Example of smoothing an active stereo scan of a styrofoam cup.	12/18/2004
	Movie: Example of smoothing and decimation of a laser scan of the head of Venus.	12/18/2004
	Movie: Example of smoothing and decimation of a laser scan of a toy monster.	12/18/2004
	Movie: Example of smoothing and decimation of a laser scan of a plaster cast of teeth.	12/18/2004
	Figure: Comparison of the present work to the QSLIM algorithm.	10/1/2005
	Figure: Example of smoothing and decimation.	10/1/2005
	Figure: Example of smoothing on a classic(al) dataset.	10/1/2005
	Figure: Comparison with Jones et al. [SIGGRAPH 2003]	10/1/2005
	Figure: Comparison with Fleishman et al. [SIGGRAPH 2003]	10/1/2005
	Figure: Example of decimation of the David model.	10/1/2005
	Figure: Comparison of various possible surface prior potentials.	10/1/2005
	Figure: Example of decimation that respects animations.	10/1/2005