A fast multi-resolution differential evolution method for multimodal image registration

A new method of finding optimal variables using self-adaptive differential evolution derived multi-modal image registration is proposed. The method features multi-resolution images achieved through down-sampling at different stages of the transformation model variable optimization process. The performance of the method is illustrated by registering magnetic resonance images to histological sections. We provide an investigation of the impact of the down-sampling factor and convergence criteria. Our findings show that given appropriately down-sampled images, and by using the proposed approach, it is possible to find optimal registration transformation model variables faster than if appropriate down-sampling is not used.