ONE-DIMENSIONAL LOGARITHMIC HARMONIC SYNTHETIC DISCRIMINANT FUNCTION FILTERS FOR SHIFT-INVARIANT, SCALE-INVARIANT, AND PROJECTION-INVARIANTPATTERN-RECOGNITION

We introduce a new approach for shift-, scale-, and projection-invaria nt pattern recognition that combines the harmonic expansion and the sy nthetic discriminant function approaches by use of a synthetic discrim inant function filter with equal-order one-dimensional logarithmic har monic components. Because projection invariance in one direction is gu aranteed by the harmonics, the required number of training images is m uch fewer than with classical synthetic discriminant function filters. (C) 1998 Optical Society of America.