Hierarchical vs. simultaneous multiresolution strategies for nonrigid image registration

Authors:
Wei Sun;Wiro J. Niessen;Stefan Klein
Affiliations:
Biomedical Imaging Group Rotterdam, Departments of Radiology and Medical Informatics, Erasmus MC, Rotterdam, The Netherlands;Biomedical Imaging Group Rotterdam, Departments of Radiology and Medical Informatics, Erasmus MC, Rotterdam, The Netherlands, Department of Image Science and Technology, Faculty of Applied Science ...;Biomedical Imaging Group Rotterdam, Departments of Radiology and Medical Informatics, Erasmus MC, Rotterdam, The Netherlands
Venue:
WBIR'12 Proceedings of the 5th international conference on Biomedical Image Registration
Year:
2012

Citing 6
Cited 0

Adaptive Stochastic Gradient Descent Optimisation for Image Registration

International Journal of Computer Vision
Fast image registration by hierarchical soft correspondence detection

Pattern Recognition
Optimal discrete multi-resolution deformable image registration

ISBI'09 Proceedings of the Sixth IEEE international conference on Symposium on Biomedical Imaging: From Nano to Macro
Topology preserving deformable image matching using constrained hierarchical parametric models

IEEE Transactions on Image Processing
Fast parametric elastic image registration

IEEE Transactions on Image Processing
Multiresolution registration of remote sensing imagery by optimization of mutual information using a stochastic gradient

IEEE Transactions on Image Processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

Nonrigid image registration algorithms commonly employ multiresolution strategies, both for the image and the transformation model. Usually a hierarchical approach is chosen: the algorithm starts on a level with reduced complexity, e.g. a smoothed and downsampled version of the input images, and with a limited number of degrees of freedom for the transformation. Gradually the level of complexity is increased until the original, non-smoothed images are used, and the transformation model has the highest degrees of freedom. In this study, we define two alternative approaches in which low- and high-resolution levels are considered simultaneously. An extensive experimental comparison study is performed, evaluating all possible combinations of multiresolution schemes for image data and transformation model. Publicly available CT lung data, with annotated landmarks, are used to quantify registration accuracy. It is shown that simultaneous multiresolution strategies can lead to more accurate registration.