Table 12. Registration failure rates for sub-shape registration. (Experiment 6).

Alg.	Failure Rate (%) by Test Case
	1	2	3	4	5	6	7
ICP	15.0	10.7	17.3	13.7	14.7	18.7	16.3
IMLP-CP	4.7	2	5.3	4.3	4.3	4.3	4.0
IMLP-MD	6.0	3.3	7.3	5.3	7.0	6.7	5.3
GICP	6.0	4.3	8.3	6.3	6.0	5.3	4.7
CPD	0.0	0.0	0.0	0.0	0.0	0.3	0.3
IMLP	6.0	3.0	7.0	5.0	6.0	6.3	5.0

Source shapes were randomly generated from from a mesh model of a human femur (Fig. 1B), misaligned by [10, 20] mm / degrees, and registered back to a point-cloud representation of the mesh. The test cases represent different noise models used to generate noise on the source shape (Table 11). For each test case, 300 randomized trials were conducted with the percent of unsuccessful registrations (TRE > 10 mm) being shown in the table. The proposed IMLP algorithm was evaluated relative to standard ICP [1], GICP [11], and CPD [20], as well as relative to near-comparisons of GTLS-ICP [10] and A-ICP [12] using the two IMLP variants IMLP-CP and IMLP-MD, which modify IMLP’s most-likely match criteria to that of closest-point and Mahalanobis-distance matching, respectively.