Abstract

Two-dimensional images with twofold symmetry such as straight lines, ellipses, and circles form an important class of objects in pattern recognition. In this paper we propose and demonstrate a novel optical method based on autoconvolution (or self-convolution) using photorefractive four-wave mixing for the detection of such objects. The 2-D spatial convolution of an image with itelf is equivalent to the correlation of the object with its 180° rotated version. For objects with twofold symmetry, the rotated version is identical to the original object and hence a strong correlation peak results. The basic idea of spatial convolution/correlation using photorefractive four-wave mixing in the Fourier plane had long been demonstrated.¹ The novel feature of our method is the capability to perform autoconvolution from a single input pattern.

© 1989 Optical Society of America