Image Registration is widely used in medical image analysis for determining the geometrical transformation, or mapping, relating two or more images. These images can be of the same subject using different modalities, or the same modality at different time points, or of different subjects. There is a large literature on medical image registration, which has become a core technology in a number of medical imaging applications. Current registration algoritms that can determine non-rigid mappings are extremely computationally demanding, with run times of 10s of minutes or hours on desktop computers. Despite the large amount of research activity in this field, there has been up to now very few comparitive evaluations of these algorithms on reference datasets.