Abstract

Geometric invariants are shape desrriptors that remain unchanged under geometric transformations such as projection, or change of the viewpoint. In [,?I we devc’loped a new method of obtaining local projer-tive antl affine invariants for a general curve without any correspondences. Being local, the invariants are much Ii ss sensitive to occlusion than global invariants. The iniiariants computation is based on a cunonicul method This consists of defining a canonical coo7.-dinate system using intrinsic properties of the shapc, independently of the given coordinate system. Since this canonical system is indept-ndent of the oriyinul one, it is invariant and all quantities defined in it arc invariant. Here we present a furth.er developnient of the method to obtain local semi-invariants, thud is lo-id rnvtrriunts for curves with known correspondencrs. Several conjigurations are treated: curves with knoion correspondences of one or two feature points or lines. 1

Citations

5	Similarity extraction and modeling - Gool, Wagemans, et al. - 1990
3	High Order Differentiation Filters that Work - Weiss - 1991
1	mehhods for det,erinining projective invariant,s in imagery”, CVG‘IP:Ilr - “General
1	Noise resistant, projectjive and affine invariants - Weiss - 1992