VocMatch: Efficient Multiview Correspondence for Structure from Motion
M. Havlena, K. Schindler: VocMatch: Efficient Multiview Correspondence for Structure from Motion (Downloadsupplementary material (ZIP, 17 MB)vertical_align_bottom), 13th European Conference on Computer Vision (ECCV), Zurich, Switzerland, 2014.
Abstract
Feature matching between pairs of images is a main bottleneck of structure-from-motion computation from large, unordered image sets. We propose an efficient way to establish point correspondences between all pairs of images in a dataset, without having to test each individual pair. The principal message of this paper is that, given a sufficiently large visual vocabulary, feature matching can be cast as image indexing, subject to the additional constraints that index words must be rare in the database and unique in each image. We demonstrate that the proposed matching method, in conjunction with a standard inverted file, is 2-3 orders of magnitude faster than conventional pairwise matching. The proposed vocabulary-based matching has been integrated into a standard SfM pipeline, and delivers results similar to those of the conventional method in much less time.
Source Code
The source code related to the paper is available here:
Downloadvocmatch-1.0.zip (ZIP, 11 KB)vertical_align_bottom
Recommended Configuration
- Intel Core-i7, 24+GB RAM, SSD
- MATLAB R2014a, Linux 64bit
Dependencies
- FLANN 1.8.4 (with OpenMP support)
external pagehttp://www.cs.ubc.ca/research/flann/call_made - Hessian affine keypoint extractor (MEX version)
external pagehttp://cmp.felk.cvut.cz/~qqmikula/publications/ijcv2012/call_made - Fine Vocabulary in FLANN 1.8.4 format
external pagehttp://ptak.felk.cvut.cz/personal/havlem1/voca/call_made - Bundler (run yourself afterwards to reconstruct 3D models)
external pagehttp://www.cs.cornell.edu/~snavely/bundler/call_made
Questions related to the code may be directed to .