Anisotropic ICP
About the A-ICP
Since its introduction in the early 1990s, the Iterative Closest Point (ICP) algorithm has become one of the most well-known methods for geometric alignment of 3D models. Given two roughly aligned shapes represented by two point sets, the algorithm iteratively establishes point correspondences given the current alignment of the data and computes a rigid transformation accordingly. From a statistical point of view, however, it implicitly assumes that the points are observed with isotropic Gaussian noise. In this project, we present the first variant of the ICP that accounts for anisotropic localization uncertainty in both input point sets as well as in both steps of the algorithm [1]. The localization error associated with a point is integrated into the algorithm via a covariance matrix representing a zero-mean Gaussian distribution. Depending on the application, the latter can be defined by the user or derived directly from the data [1-2]. We refer to this generalized ICP variant as anisotropic ICP (A-ICP).
The A-ICP algorithm was implemented as plugin within the MITK framework. An evaluation on publicly available surface meshes as well as on a set of meshes extracted from medical imaging data shows a dramatic increase in accuracy compared to the original ICP, especially in the case of partial surface registration [1]. Furthermore, an extension for registration of partially overlapping surfaces, referred to as trimmed A-ICP, was used in a mobile augmented reality application [3-4].
Downloads
On this page we provide a software package containing our implementation of the A-ICP algorithm. A brief user guide helps applying the algorithm to a set of provided test data.
- Software Package Windows 32 Bit
- Software Package Windows 64 Bit
- Software Package Linux 64 bit (Ubuntu 12.04 or higher; requires liblapack-dev; to install liblapack-dev open a terminal and type: "sudo apt-get install liblapack-dev")
- User Guide
- Testdata Package
References
[1] L. Maier-Hein, A. M. Franz, T. dos Santos, M. Schmidt, H.-P. Meinzer,and J. M. Fitzpatrick. Convergent Iterative Closest-Point Algorithm to Accomodate Anisotropic and Inhomogenous Localization Error. IEEE T Pattern Anal, 34(8): 1520–1532, 2012.
[2] L. Maier-Hein, M. Schmidt, A. M. Franz, T. R. dos Santos, B. Jähne, J. M. Fitzpatrick, and H.-P. Meinzer. Accounting for anisotropic noise in fine registration of Time-of-Flight range data with high-resolution surface data. In Medical Image Computing and Computer-Assisted Intervention - MICCAI 2010 (1), 6361, pages 251–258, 2010.
[3] L. Maier-Hein, A. M. Franz, M. Fangerau, A. Seitel, S. Mersmann, T. Kilgus, A. Groch, K. Yung, T. R. dos Santos, and H.-P. Meinzer. Towards Mobile Augmented Reality for On-Patient Visualization of Medical Images. In Bildverarbeitung für die Medizin 2011, 389–393, Springer, 2011.
[4] T. Kilgus, A. M. Franz, A. Seitel, K. März, L. Bartha, M. Fangerau, S. Mersmann, A. Groch, H.-P. Meinzer, and L. Maier-Hein. Registration of partially overlapping surfaces for range image based augmented reality on mobile devices. In SPIE Medical Imaging 2012: Image-Guided Procedures, Robotic Interventions, and Modeling, 83160T (8 pages)