IMLP runtime comparison using different PD-tree bounding methods for computing the most-likely matches.

Average runtimes are reported for the proposed IMLP algorithm over the 300 registration trials of Experiment 4B, Test Case 1, which involves registering 100 random samples to a point-cloud representation of a hip model (Fig. 1A). Runtimes were recorded for the naive exhaustive search and for the proposed PD-tree search strategy comparing two of the proposed PD-tree bounding methods: the spherical (14) and simple ellipsoidal (15) bounds. The compact ellipsoidal bound (17) was not evaluated. IMLP runtime comparison using different PD-tree bounding methods for computing the most-likely matches.