Quasi-invariant parameterisations and their applications in computer vision

In this paper, we show that there exist quasi-invariant parameterisations w hich are not exactly invariant but approximately invariant under group tran sformations and do not require high order derivatives. The affine quasi-inv ariant parameterisation is investigated in more detail and exploited for de fining general affine semi-local invariants from second order derivatives o nly. The new invariants are implemented and used for matching curve segment s under general affine motions and extracting symmetry axes of objects with 3D bilateral symmetry.