The internal and relative orientations of one or more cameras can be obtained by tracking a single point from each of two or more cameras. Thus, for example, a stereo rig can be internally and relatively calibrated by walking around in front of it, or a single camera can be internally calibrated this way using a second camera for data collection.
The implications of this fact are important for bringing 3-D computer vision systems out of the laboratory and into practical use. Cumbersome calibration objects and all but the simplest image processing are not required. The accuracy of calibration that can be achieved has no theoretical limit since an arbitrarily large amount of data can be collected. Also, our calibration procedure addresses problems that are difficult or impossible for traditional calibration techniques, including wide-baseline stereo.
We present a solution technique for self-calibration from one tracked point and verify its viability on synthetic and real stereo rigs.