Abstract

Recent developments in optical computing have the promise of higher flexibility and easier implementation of signal processing functions such as transforms.¹ In the past, phase-only and two-bit phase-only matched filtering showed good results.² However, the nonlinear operations involved were restricted to the transfer function of the matched filter. We show that considerably better accuracy is possible by extending the nonlinear operations to the spectrum of the signal itself.

© 1987 Optical Society of America