Abstract

The use of synthetic discriminant function filters (SDFs) has long been proposed as a means of achieving distortion-invariant pattern recognition in optical correlators. Various design algorithms have been developed toward this end, including binary phase-only SDFs that can be readily implemented with current types of spatial light modulator. Recently, optical filters have been developed in which certain of the binary filter’s pixels are blocked to achieve a ternary state filter.¹ These filters have been demonstrated to provide better correlation discrimination between the in-class object and an out-of-class object. In this work we design ternary phase and amplitude synthetic discriminant function filters in simulation with an algorithm that allows equal correlation peaks to be obtained for in-class objects while suppressing the responses from out-of-class objects, relative to the discrimination ratios achieved with binary phase-only SDFs. We investigate issues pertinent to practical systems’ use including variations in the design procedure, the degree of discrimination improvement afforded, response level of images between the training set images, maximum distortion range possible and signal-to-noise considerations.

© 1991 Optical Society of America