Table 8. Registration failure rates for registering a point-cloud target shape. (Experiment 4B).

Alg.	Failure Rate (%) by Test Case
	1	2	3	4	5	6	7	8	9
ICP	0.3	0.7	0.3	0.7	0.3	2.0	0.7	0.3	0.3
IMLP-CP	0.3	0.7	0.3	0.7	0.3	2.0	0.7	0.3	0.3
IMLP-MD	0.3	0.7	0.3	0.7	0.3	2.0	1.0	0.3	0.3
GICP	1.3	1.3	0.7	2.7	3.7	2.0	2.0	1.3	1.7
CPD	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
IMLP	0.3	0.7	0.3	0.7	0.3	2.0	1.0	0.3	0.3

Source shapes were randomly generated from a mesh model of a human hip (Fig. 1A), misaligned by [15, 30] mm / degrees in (Experiment 4A) and [30, 60] mm / degrees in (Experiment 4B), 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 4). For each test case, 300 randomized trials were conducted, with the percent of unsuccessful registrations (TRE > 10 mm) being shown in this 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 IMLP-CP and IMLP-MD, which modify IMLP’s most-likely match criteria to that of closest-point and Mahalanobis-distance matching, respectively. Failure rates for Experiment 4A (which are not shown in the table) were 0% for all algorithms and test cases, except for test case 2, where standard ICP incurred one registration failure.