open3DCV

open3DCV is a Multiple View Geometry library developed by Kai Wu for efficient and reliable implementation of MVG algorithms. The goal is to provide a lightweight implementation with minimum dependencies. This library implements many widely used vision algorithms, which includes

  • camera models
  • keypoint detection and description extraction
  • feature matching
  • pose estimation
    • seven-point algorithm for Fundamental matrix
    • eight-point algorithm for Fundamental matrix
    • five-point algoirthm for Essential matrix
  • N-view triangulation
  • robust estimation using RANSAC

Implementation

The implementation of various algorithms have been organized into a list of blog posts. Please visit the page of open3DCV implementation tutorials for more details.

Download

The source code and documentation are available in Github

Acknowledgements

open3DCV is one of the outcomes of my Master’s work at the HCT lab of University of British Columbia.