A non-zero order joint transform correlator for space-variant pattern recognition

We propose a space-variant correlator using a joint transform correlator (JTC) architecture. The reference and target images of a JTC are modulated by a random phase function which is the sample of a stationary white process ergodic in mean and variance. The degree of shift-variance can be varied by changing the variance of the random phase function. The joint power spectrum (JPS) is a sinusoidally varying function with a spatially nonvarying average value. This property is advantageous in optical implementations when the JPS is captured by devices such as CCD with a limited dynamic range. The zero-order and the correlations approach delta functions in the correlation plane. By subtracting a constant value equal to its spatial average from the JPS, the undesirable zero-order can be eliminated. (C) 1999 Elsevier Science B.V. All rights reserved.