Image Processing Reference
In-Depth Information
Surface registration by
markers guided nonrigid
iterative closest points
Dominik Spinczyk Faculty of Biomedical Engineering, Silesian University of Technology, Zabrze, Poland
The problem of matching irregular surfaces was tested with additional markers as landmarks for the ex-
tension of the nonrigid iterative closest points algorithm. The general idea of presented approach was
to take into account knowledge about markers' positions not only in computing transformation phase
but also in finding correspondence phase in every algorithm's iteration. Four variants of retrieving cor-
respondence were implemented and compared: the Euclidean distance, normal vectors with initial ri-
gid registration, static and dynamic markers vectors. To evaluate different manner of computing corres-
pondence the average correspondence assignment error of points nearest to the markers and the number
of correspondences for every target points were defined. The presented approach was evaluated using
abdominal surfaces data set consisting of captured clouds of points during free breathing of six volun-
teers. The modifications significantly improved results. To make the proposed changes more universal
k-nearest neighbor method and radius constraint could be used.
Markers guided geometry-based registration
Finding correspondences
Nonrigid iterative closest points
Surface registration quality
Surface registration
The study was supported by National Science Center, Poland, Grant No UMO-2-12/05/B/ST7/
Search WWH ::

Custom Search