Abstract

This paper presents a modified version of the Stanford optical vector-matrix multiplier1 which is able to perform matrix-matrix multiplication. In the Stanford processor, an input row vector is multiplied by a matrix mask to give a resulting output column vector. If two row vectors are present in the input plane, the two output column vectors are spatially superimposed in the output plane. This makes it impossible to distinguish the two column vectors. However, if a linear phase mask (which varies in phase along the length of the row vector) is placed over one of the input row vectors, its output column vector is displaced to the side of the other column vector.

© 1987 Optical Society of America