A hierarchy of invariant representations of 3D shape

The representation of geometrical 3D shape is discussed in this paper. A hierarchy of representations, which are invariant under weak perspective to either rigid or affine 3D transformations is described. These representations can be computed efficiently from a sequence of images with a linear and incremental algorithm. This hierarchy provides the means for adding a controlled amount of quantitative information to a symbolic representation. It also provides a library of representations, which may be used for different tasks (or purposes).<>