| |
| | [No title] |
 | | \vspace{-0.2in} \subsection{Geodesic Interpolating Splines} \vspace{-0.1in} The usual approach to constructing a large-deformation diffeomorphism is to consider such a deformation as an infinite sequence of infinitesimal deformations~\cite{CY,vsmooth,JM,Trouve}; that is, we have an infinite sequence of the spline-part generated by the Green's function $G$, and an infinite sequence of infinitesimal affine transformations. |
 | | It can be seen that the warps are indeed diffeomorphic, and appear to be very smooth~-- each of the brain slices still looks biologically plausible, which would not be true had a simpler scheme been used. |
 | | This is because, in the limit where the lines become infinitely densely sampled, it is actually impossible to construct a diffeomorphism for which the lines cross, which would mean that the geodesic distance for the illegal shapes would approach infinity as the sampling density increased. |
| www2.wiau.man.ac.uk /caws/Conferences/10/proceedings/8/papers/40/IPMI_paper%2Etex (3453 words) |
|