Enhanced affine invariant shape description by ESPRIT

Affine transform best describes the shape change of an image that is viewed in different angles. Affine invariant shape descriptor provides us a means to recognize the object in an image regardless of shape change of the object. In this paper, three affine invariant shape descriptors are presented, which are all based on Fourier transform. To suppress the inherent shortage of Fourier transform, estimation of signal parameters via rotational invariance techniques (ESPRIT) is adopted to improve the discriminability of shape descriptor.