Occluding Contours for Multi-View Stereo

Qi Shan, Brian Curless, Yasutaka Furukawa, Carlos Hernandez, and Steven M. Seitz

CVPR 2014 Paper [pdf 28M]
CVPR 2014 Poster [pdf 25M]
Supplementary video [YouTube] [480p 30M] [720p 150M]

Abstract: This paper leverages occluding contours (aka “internal silhouettes”) to improve the performance of multi-view stereo methods. The contributions are 1) a new technique to identify free-space regions arising from occluding contours, and 2) a new approach for incorporating the resulting free-space constraints into Poisson surface reconstruction. The proposed approach outperforms state of the art MVS techniques for challenging Internet datasets, yielding dramatic quality improvements both around object contours and in surface detail.

Reconstructing 3D models from Internet photos with cleaner silhouettes and improved surface details

PMVS+Poisson Proposed approach

Key Ideas:

(1) Densifying depth maps using occluding contours, and then computing a free space volume
One input color image Initial depth map based on visible PMVS points Estimated dense depth map
(2) Augmenting the PMVS point set using the dense depth maps
Direct output of PMVS Augmented point set
(3) Incorporating free space constraints in mesh reconstruction
PMVS+Poisson PMVS+Poisson colored Proposed approach  with color

*** End ***