A Comparison and Evaluation of Multi-View Stereo Reconstruction Algorithms Steve Seitz and Brian Curless, University of Washington, Seattle, WA James Diebel, Stanford University, Palo Alto, CA Daniel Scharstein, Middlebury College, Middlebury, VT Richard Szeliski, Microsoft Research, Redmond, WA

Abstract— This paper presents a quantitative comparison of several multi-view stereo reconstruction algorithms. Until now, the lack of suitable calibrated multi-view image datasets with known ground truth (3D shape models) has prevented such direct comparisons. In this paper, we first survey multi-view stereo algorithms and compare them qualitatively using a taxonomy that differentiates their key properties. We then describe our process for acquiring and calibrating multiview image datasets with high-accuracy ground truth and introduce our evaluation methodology. Finally, we present the results of our quantitative comparison of state-of-the-art multi-view stereo reconstruction algorithms on six benchmark datasets. The datasets, evaluation details, and instructions for submitting new models are available online at Multi-View Stereo Evaluation Homepage (http://vision.middlebury.edu/mview).

This paper was presented at the IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR06).

	Paper: CVPR06 paper [PDF] [Bibtex].	4/2/2006
	Presentation: the data collection and processing in PPT format but packaged as a ZIP file that contains all the animations.	10/1/2005
	Movie: image sequence of the bird statue, in MPG4 format.	9/8/2005
	Movie: image sequence of the statue of a boy, in MPG4 format.	9/8/2005
	Movie: image sequence of the dinosaur statue, in MPG4 format.	9/8/2005
	Movie: image sequence of the statue of two dogs in a basket, in MPG4 format.	9/8/2005
	Movie: image sequence of the Greek temple model, in MPG4 format.	9/8/2005
	Movie: image sequence of the pine cone, in MPG4 format.	9/8/2006
	Movie: 3D model of Greek temple, as scanned by Cyberware Model 15 scanner.	9/8/2005
	Movie: 3D model of Greek temple, texture mapped by images and relit.	9/8/2005