To obtain an invariant representation with respect to the global position and orientation of the considered structure, we can express the configuration of all frames relative to one frame (the basis). For efficiency, we store this representation in a hash-table and for correctness, we include the uncertainty of each invariant.
As only a few frames are in the same configuration in the two images, the one chosen as the basis may not be present in the other image. The preprocessing step is thus repeated with each frame as the basis.
[ Back to substructure
matching with frames ] 