Abstract

A joint transform correlator (JTC) is suitable for real-time matching and tracking operations without the fabrication of a complex filter. However, a classical JTC suffers from low correlation discrimination, large correlation sidelobes, and low diffraction efficiency. A binary JTC may overcome these problems by using nonlinearity in the Fourier plane to binarize the joint power spectrum (JPS). However, JPS binarization is a computation intensive process and may slow down the processing speed of the correlator. In filter based pattern recognition, an amplitude-modulated phase-only filter is found to yield significantly better correlation discrimination than a phase-only filter.¹ Applying this technique in JTC, an amplitude-modulated JTC (AMJTC) is found to yield better correlation discrimination than a binary JTC for noise-free inputs.² However, the main difficulty of this particular AMJTC is the selection of the threshold value used to calculate an amplitude-modulated filter corresponding to the reference. In this paper, we attempt to overcome this problem and study the performance of the AMJTC for noisy input images via computer simulation.

© 1992 Optical Society of America