“…Practically, however, most registration methods in use have tended to use asymmetric cost functions that lead to a solution that is not inverse consistent when the template and the target images being registered are interchanged. Several papers have recently attempted to make the registration symmetric or inverse consistent [1]- [3], [8], [10], [11], [15], [17], [18], [20]- [22], [24], [25]. Here, we present two novel cost functions in the large deformation image matching framework that are symmetric with respect to the choice of template and target images being matched.…”