ECM versus ICP for point registration

Xie, Weiguo; Nolte, Lutz-Peter; Zheng, Guoyan (2011). ECM versus ICP for point registration. In: The 33rd Annual International Conference of the IEEE EMBS (EMBC 2011). IEEE Engineering in Medicine and Biology Society conference proceedings: Vol. 2011 (pp. 2131-2135). Vancouver, Canada: IEEE 10.1109/IEMBS.2011.6090398

Preview

Text 06090398.pdf - Published Version Available under License Publisher holds Copyright. © 2011 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. Download (834kB) | Preview

Iterative Closest Point (ICP) is a widely exploited method for point registration that is based on binary point-to-point assignments, whereas the Expectation Conditional Maximization (ECM) algorithm tries to solve the problem of point registration within the framework of maximum likelihood with point-to-cluster matching. In this paper, by fulfilling the implementation of both algorithms as well as conducting experiments in a scenario where dozens of model points must be registered with thousands of observation points on a pelvis model, we investigated and compared the performance (e.g. accuracy and robustness) of both ICP and ECM for point registration in cases without noise and with Gaussian white noise. The experiment results reveal that the ECM method is much less sensitive to initialization and is able to achieve more consistent estimations of the transformation parameters than the ICP algorithm, since the latter easily sinks into local minima and leads to quite different registration results with respect to different initializations. Both algorithms can reach the high registration accuracy at the same level, however, the ICP method usually requires an appropriate initialization to converge globally. In the presence of Gaussian white noise, it is observed in experiments that ECM is less efficient but more robust than ICP.

Interest & Impact

Downloads

525 since deposited on 04 Oct 2013
41 in the past 12 months

Citations

5 Citations in Web of Science ®
6 Citations in Scopus

Search

in Google Scholar™

Services

Actions (login required)

Edit item

Actions (login required)

Edit item