Abstract

Design of smart filters for programmable hybrid optical/digital correlators generally requires execution of sophisticated algorithms in a time-consuming computer program. The performance of the actual correlator hardware is often disappointing, because overly simplified and idealized models of the optical components are usually assumed during filter design. We suggest a technique in which both a computer and the front half of the hybrid correlator itself are combined in a joint transform configuration to provide accurate feedback about the optical system during filter design. Differences between the phase and amplitude of a reference image and its various scaled and rotated distortions are detected in the filter plane and decoded from the joint transform fringe patterns using spatially local operations. The differences can be applied to design filters that give more predictable correlator performance.

© 1988 Optical Society of America