Non-rigid motion estimation from hierarchical adaptive local affine registrations

Non-rigid image registrations have been widely employed in medical imaging to estimate the complex tissue deformations such as those caused by respiratory motion. To reduce the computational burden of non-rigid registration techniques, alternative approaches have been proposed which decompose the non-rigid registration problem into a combination of multiple more simple registration problems, such as locally affine components. For the EMPIRE10 challenge we apply a modified version of a local affine registration algorithm, where all local affine components are embedded in an adaptive hierarchical structure.