1: \begin{abstract}

2:   We discretize a cost functional for image registration

3:   problems by deriving Taylor expansions for the matching term. Minima of the

4:   discretized cost functionals can be computed with no spatial discretization error,

5:   and the optimal solutions are equivalent to minimal energy curves in the

6:   space of $k$-jets. We show that the solutions convergence to 

7:   optimal solutions of the original cost functional as the number of particles

8:   increases with a convergence rate of $O(h^{d+k})$

9:   where $h$ is a resolution parameter. The effect of this approach over traditional

10:   particle methods is illustrated on synthetic examples and real images.

11: \end{abstract}

12: