The first step is building of shape models, which use a finite number of modes for representation. Shape models that are built from the outlines of objects in images enable those images to be brought into a state of alignment. We can warp the shapes within an image to match the mean shape. By interpolation, we can extend this warp to the entire interior of the object. This means that corresponding parts of shapes in those images will be easy to identify and then use, e.g. in order to sample intensities.