Joint Fourier transform correlation is an analog optical signal processing technique capable of comparing several two-dimensional images. We show that selected spectral domain binary thresholds, obtained through simulation, significantly improve the performance of an optical implementation of the joint transform correlator. In doing so we establish empirically the existence of well defined optimal uniform thresholds for the image classes tested. We also investigate the discrimination capabilities of the linear joint transform correlator, both by simulation and optically, in multiple input environments.
Keywords: optical pattern recognition, binary joint transform correlator, spectral domain processing, optimal thresholding.