Multimodal image registration by particle filtering: Evaluation and new results

This paper presents a study of parametric multimodal image registration based on particle filtering. We show that this methodology offers high performance over traditional approaches based on mutual information and gradient descendent optimization. The evaluation was performed with a set of medical images, and the tests include evaluations under different noise conditions and partial data, and a comparison with a state of the art technique. Additionally, the proposed methodology was improved by using an extended model in the stochastic search. Finally, a multithreads implementation by multi-core processing was developed that reduced the computational time in the registration task. According to the evaluation, the parametric registration based on particle filtering shows accuracy, robustness and high computational speed in the parameters estimation of multimodal image registration, and it could be considered as an efficient tool in medical imaging applications. © 2003-2012 IEEE.

Image registration; medical imaging; particle filters; similarity metrics Imaging applications; Multi-core processing; Multimodal image registration; Parameters estimation; Particle filter; Similarity metrics; State-of-the-art techniques; Traditional approaches; Image registration; Medical imaging; Monte Carlo methods; Optimization; Stochastic models; Signal filtering and prediction

VIVO