Abstract

There are two major drawbacks in a conventional joint transform correlator (JTC) for multi-target pattern recognition, i.e., the broad correlation profiles and the low energy efficiency. We have shown recently, by using an inverse reference power spectrum in the frequency domain for a JTC, that the output autocorrelation profiles can be sharpened dramatically. However, this technique was mainly used in a single reference function. For multi-target detection, the joint transform spectra should be compensated by each inverse reference spectrum. Since it is inefficient to perform this task in the frequency domain, in this paper we have used a spatially separable synthesis method to overcome this impediment, by which one-step peak-sharpening compensation can be achieved. To suppress the adverse spectral contents and remove some unwanted diffractions, frequency domain manipulation is also introduced. The effects due to the spectral dynamic range limit, the noise, and cluttering have been investigated. As compared with the conventional JTC, we have shown that the proposed system offers higher accuracy of detection, higher reliability, and higher diffraction efficiency.

© 1992 Optical Society of America