Flexible models or deformable templates are now widely used to aid image interpretation. Such models usually have a number of parameters to control the shape and pose of all or part of the model. For instance Yuille et al [4] and Lipson et al [5] use `hand-crafted' models of faces and transaxial slices through vertebrae. Scott [6] and Staib and Duncan [7] use closed contour shape models, based on expansions of trigonometric functions, for the interpretation of medical images. Pentland and Sclaroff [8] and Nastar and Ayache [9] work with finite element models of flexible objects. Grenander, Chow and Keenan [10] and Mardia, Kent and Walder [11] describe statistical models of shape. Bookstein [12] has applied statistical techniques to learn relationships between shape and other variables for morphometric analysis.