1: \begin{abstract}

2: The Procrustes distance is used to quantify the similarity or dissimilarity of (3-dimensional)

3: shapes, and extensively used in biological morphometrics. Typically each (normalized) shape

4: is represented by $N$ landmark points, chosen to be homologous (i.e. corresponding to each other), as much as possible, and the Procrustes

5: distance is then computed as $\inf_{R}\,\sum_{j=1}^N\, \|Rx_j-x'_j\|^2$, where the minimization is over all Euclidean transformations, and the correspondences $x_j \leftrightarrow x'_j$

6: are picked in an optimal way. \\

7: This (discrete)

8: Procrustes distance is easy to compute but 

9: has drawbacks -- representeding a shape by only a finite number of

10: points, which may fail to capture all the geometric aspects of interest; a need has been expressed

11: for alternatives that

12: are still computationally tractable.

13: We propose in this paper the concept of 

14: %{\large

15: {\bf {continuous Procrustes distance}}, and prove that it provides a true metric

16: for two-dimensional surfaces embedded in three dimensions. The continuous Procrustes 

17: distance leads to a hard optimization

18: problem over the group of area-preserving diffeomorphisms. One of the core 

19: observations of our paper is that for small continuous Procrustes distances, 

20: the global optimum of the Procrustes distance can be uniformly approximated 

21: by a conformal map. This observation leads to an

22: efficient algorithm to calculate approximations to this new distance.

23: \end{abstract}