Image recognition by affine moment invariants in Hartley transform domains

The determination of affine invariant characteristics is an important problem in pattern recognition. Existing methods to obtain affine invariants are mainly based on moments in the spatial domain. This paper introduces affine moment invariants in Hartley transform domain. The magnitude of the Hartley transform is reconstructed by thresholding. Several affine moment invariants are obtained from the reconstructed image in Hartley transform domain. These affine invariant features can be used to recognize the distorted images in spatial domain. It is found that affine invariants in the Hartley transform domain are robust to noise and the recognition rate is increased. Experimental results are presented to illustrate the performance of the invariants for affine deformed images.