Home

• Introduction

  PMVS is a multi-view stereo software that takes a set of images and camera parameters, then reconstructs 3D structure of an object or a scene visible in the images. Only rigid structure is reconstructed, in other words, the software automatically ignores non-rigid objects such as pedestrians in front of a building (See examples in our gallery). The software outputs a set of oriented points instead of a polygonal (or a mesh) model, where both the 3D coordinate and the surface normal are estimated at each oriented point. Michael Misha Kazhdan provides a couple of programs that convert a set of oriented points into a surface model, and you can download his software from his website.

  Our software can be downloaded (after signing the agreement form and waiting for the acknowledgement) from here. We provide binary versions of our software for Windows, which were compiled on a 32bit Windows XP machine. Note that this is a multi-view stereo software that assumes that cameras are already calibrated. You have to calibrate your cameras before using our software.
  
  Quantitative evaluations of our software is found at the Multi-View Stereo Evaluation website. As of 9/25/2007, our software has the best results (both in terms of accuracy and completeness) for 4 out of their 6 data sets. The algorithm implemented in the software is described in our CVPR 2007 paper. You can find our 3D photography data sets here and there. Note that many data sets in the examples do not belong to us, and hence we cannot distribute them.
  
  
  

• Contacts

  For comments or bug-reports, you can write to Yasutaka Furukawa (furukawa -at- cs.washington.edu). Please use "multi view stereo" as a subject of an email. However, we do not provide any support or help regarding the software beyond the documentation, as we are getting many emails and will not have time to answer each one of them.
  

• Reconstruction Examples (more in our gallery)

  
  

• Acknowledgements
  This work was supported in part by the National Science Foundation under grant IIS-0535152. We thank S. Seitz, B. Curless, J. Diebel, D. Scharstein, and R. Szeliski for the temple and dino datasets and evaluations, Carlos Hernández Esteban, F. Schmitt, and the Museum of Cherbourg for polynesian, S. Sullivan, A. Suter, and Industrial Light and Magic for face, face-2, body, steps, and wall, C. Strecha for city-hall and brussels, and finally J. Blumenfeld and S. R. Leigh for the skull data set. Our software uses WNLIB, developed by Will Naylor and Bill Chapman, for an optimization method.

Back to my home page -- Last updated on 2008/08/25.

.