Robust and Efficient 3-D Reconstruction by Self-Calibration.
Hanno Ackermann, Kenichi Kanatani · 2007
We present a very robust and efficient scheme for 3-D reconstruction from point correspondences over multiple images taken by uncalibrated cameras. Our method is based on iterative projective reconstruction followed by Euclidean upgrading. We accelerate the time-consuming iterations for projective reconstruction by using the power method for eigenvalue computation, accelerating the power method itself, and introducing SOR for further acceleration. Using simulated and real video images, we demonstrate that our techniques dramatically speed up the computation, in some cases about 8,000 times faster. 1.