next up previous index
Next: MRF Parameter Estimation Up: Pose Computation Previous: Simultaneous Matching and Pose

5.3.3

Discussion

Minimizing the energies   derived in this section is difficult. The dimensionality of the search space is high. As a guide to the search, the energies are inefficient unless some strong prior constraints are available, such as the normal prior distribution of poses assumed in [Wells III 1991]. However, the derived models may be useful for the verification of matching and pose. Assume that pose candidates are found by using some techniques such as Hough transform or geometric indexing. The energies may be used as global cost measures for the matching and pose.