ORIENTATION AND AUTO-CALIBRATION OF IMAGE TRIPLETS AND SEQUENCES

The automatic orientation and auto-calibration of image triplets and sequences is the basis for applications in visualization but also for all tasks employing metric information from imagery. Here we propose a hierarchical approach for orientation based on image triplets which is robust in the sense that it works with one set of parameters for a larger number of different sequences, also containing mediumto wide-baseline images, which standard procedures cannot handle. For auto-calibration we have made tests using the absolute dual quadric and the stratified approach based on the modulus constraint, showing that our results correspond with given calibration data, but are not yet totally stable. Finally, we propose a simple and robust method which allows the determination of the two parameters of the principal distance in x- and y-direction for image triplets robustly and reliably.