In nonrigid image registration, B-splines are used extensively for interpolation and transformation. We propose efficient multi-dimensional algorithms for the B-spline interpolation and transformation functions and their derivatives. The algorithms are based on recursive formulations and implemented using template metaprogramming. Compared to reference implementations, we obtain an acceleration factor of 4 for interpolation and a factor of 18 for transformation. When used within a registration algorithm, total computation time reduces by a factor 1.5 to 3.5.

This presentation is part of Minisymposium “MS47 - Splines in Imaging (3 parts)”
organized by: Carolina Beccari (Dept. Mathematics, University of Bologna) , Virginie Uhlmann (EPFL, Lausanne) , Michael Unser (EPFL, Lausanne) .

Authors:

Stefan Klein (Biomedical Imaging Group Rotterdam, Erasmus MC)

Wyke Huizinga (Biomedical Imaging Group Rotterdam, Erasmus MC)

Marius Staring (Division of Image Processing (LKEB), Dept. of Radiology, Leiden University Medical Center)

Dirk Poot (Biomedical Imaging Group Rotterdam, Erasmus MC)

Keywords:

b-spline, image registration, interpolation, template metaprogramming, transformation