Orientation Constraints in Point-Based Elastic Image Registration

Landmark-Based Image Analysis

“…Generally, these attributes characterize the local orientation of the contours at the landmarks (Fornefett et al [158]- [ 160], Rohr et al [448]). As attributes we here consider orientations at point landmarks in both images to be mapped onto each other.…”

“…Generally, these attributes characterize the local orientation of the contours at the landmarks (Fornefett et al [158]- [ 160], Rohr et al [448]). While in [158]- [160] we have treated the case of isotropic landmark errors together with orientation attributes, in [448] we have extended this scheme for inclusion of anisotropic errors. Thus, we can integrate statistical as well as geometric information as additional knowledge in elastic image registration.…”

“…For additional experimental results including the 3D case, see [158]- [160], [448]. One problem with this type of approach is that the influence of incorporated orientations is rather global, i.e., image parts far away from landmarks with added orientations are generally strongly affected, which is often not desired.…”

“…Mardia et al [324], [325] have proposed an exact analytic solution based on the stochastic method of kriging. In comparison to these schemes, our approach for incorporating orientations in Fornefett et al [158]- [160] and Rohr et al [448] is based on a minimizing functional within the framework of thin-plate splines. They also proposed to incorporate curvature attributes at landmarks .…”

“…It is required that the orientation vectors are unit vectors. Whereas in [158]- [ 160] we have treated the case of isotropic landmark errors together with orientation attributes, in [448] we have extended this scheme for inclusion of anisotropic errors. Both approaches in [55] and [324], [325] deal with interpolating transformations.…”