Three-dimensional face recognition

A method of image preprocessing of three-dimensional topographical data of a manifold surface having a topography for subsequent classification, the method comprising:

• providing image preprocessing hardware;
• and
• within said image preprocessing hardware:
• placing said three-dimensional topographical data over a two-dimensional manifold to form said manifold surface having a topography,
• converting said manifold into a series of geodesic distances between pairs of points over said manifold, said geodesic distances being measured over said topography of said manifold,
• generating a matrix of geodesic distances over said manifold to selected vertices of said manifold, and
• using multi-dimensional scaling to reduce said matrix to a bending-invariant canonical representation in a low-dimensional Euclidean space, thereby to provide a representation suitable for subsequent classification.

A method of classifying images of three-dimensional bodies having surface topographies, the method comprising:

• providing image processing hardware;
• and
• within said image processing hardware:
• obtaining bending invariant representations of said three dimensional bodies as canonical form representations derived from geodesic distances between selected sample points taken from surfaces of said bodies, said geodesic distances being measured over said respective surface topographies of said three-dimensional bodies;
• from each representation deriving co-ordinates on a feature space, and
• classifying said bodies according to clustering on said feature space.