Globally Optimal Algorithms for Stratified Autocalibration

This page contains code/data/results for a stratified autocalibration method that, starting from a projective reconstruction

Step 1 (affine upgrade): performs an affine reconstruction by estimating the plane at infinity
Step 2 (metric upgrade): performs a metric reconstruction by estimating the internal parameters of the camera.

The main features of the algorithm are:

Globally optimal solution to the modulus constraints for estimating the plane at infinity.
Globally optimal solution to the infinite homography relation for estimating the dual image of absolute conic.
Incorporates a theoretically correct initial region for plane at infinity search using chirality constraints.
Enforces non-linear constraints such as positive semidefiniteness of the dual image of absolute conic.

Code:

An early version can be found [here]. Newer version coming soon!

Reference:

M.K. Chandraker, S. Agarwal, D.J. Kriegman and Serge Belongie
Globally Optimal Affine and Metric Upgrades in Stratified Autocalibration [PDF]
ICCV 2007, Rio de Janeiro, Brazil.

Last updated October 29, 2007.