Abstract

A generalized discrete 1-D optical processor with a 2-D kernel can be viewed as a matrix–vector multiplier. Since most of the data processed in optics are 2-D, we need to investigate a generalized 2-D processor which has a 4-D kernel. This processor can compute the output in one time step given the entire input data, which is in contrast with a regular systolic array which accepts time-serial input data and accumulates the output over many time steps.

© 1987 Optical Society of America