Pose Estimation Using Parametric Stereo Eigenspaces
Thomas Melzer, Michael J. Reiter · 2000
We propose new approach to constructing object models in parametric eigenspace based on stereo pairs. For each object pose, two images are taken from different viewing positions separated by a fixed baseline, and concatenated be- fore they are projected into eigenspace. Experimental results indicate that the stereo approach is superior to its conven- tional counterpart in terms of accuracy of the estimated pose parameters.