Principal Warps: Thin-Plate Splines and the Decomposition of Deformations
IEEE Transactions on Pattern Analysis and Machine Intelligence
A survey of image registration techniques
ACM Computing Surveys (CSUR)
Numerical recipes in C (2nd ed.): the art of scientific computing
Numerical recipes in C (2nd ed.): the art of scientific computing
Handbook of mathematics (3rd ed.)
Handbook of mathematics (3rd ed.)
MICCAI '00 Proceedings of the Third International Conference on Medical Image Computing and Computer-Assisted Intervention
Iconic feature based nonrigid registration: the PASHA algorithm
Computer Vision and Image Understanding - Special issue on nonrigid image registration
2-D and 3-D Image Registration: for Medical, Remote Sensing, and Industrial Applications
2-D and 3-D Image Registration: for Medical, Remote Sensing, and Industrial Applications
A Geometric Theory of Symmetric Registration
CVPRW '06 Proceedings of the 2006 Conference on Computer Vision and Pattern Recognition Workshop
Image Registration with Guaranteed Displacement Regularity
International Journal of Computer Vision
STACOM'11 Proceedings of the Second international conference on Statistical Atlases and Computational Models of the Heart: imaging and modelling challenges
Diffeomorphic cardiac motion estimation with anisotropic regularization along myofiber orientation
WBIR'12 Proceedings of the 5th international conference on Biomedical Image Registration
Hi-index | 0.00 |
This paper presents a method for constructing symmetric and transitive algorithms for registration of image sequences from image registration algorithms that do not have these two properties. The method is applicable to both rigid and nonrigid registration and it can be used with linear or periodic image sequences. The symmetry and transitivity properties are satisfied exactly (up to the machine precision), that is, they always hold regardless of the image type, quality, and the registration algorithm as long as the computed transformations are invertable. These two properties are especially important in motion tracking applications since physically incorrect deformations might be obtained if the registration algorithm is not symmetric and transitive. Themethod was tested on two sequences of cardiac magnetic resonance images using two different nonrigid image registration algorithms. It was demonstrated that the transitivity and symmetry errors of the symmetric and transitive modification of the algorithms could be made arbitrary small when the computed transformations are invertable, whereas the corresponding errors for the nonmodified algorithms were on the order of the pixel size. Furthermore, the symmetric and transitive modification of the algorithms had higher registration accuracy than the nonmodified algorithms for both image sequences.