Output details

A variational approach to video registration with subspace constraints

<23>Key success in one of the most important open problems in Computer Vision: the estimation of dense correspondences between image pixels in a video sequence in the presence of non-rigid motion. We provide a new, more efficient, robust and accurate solution which substantially outperforms state-of-the-art algorithms. Additionally, in response to the absence of ground-truth data for performance analysis of non-rigid video registration algorithms, we have released a new benchmark dataset. Invited paper, extended version of EMM-CVPR conference publication. Its application to medical imaging have resulted in an EPSRC funded collaboration with the world leading William Harvey Institute, in cardiovascular research.