The Use of Complex Spatial Frequency Filters in Correlation Processes

Abstract

A method of solving pattern recognition problems is to correlate the unknown data with the known pattern. Where the data is available in two-dimensional transparency form it is convenient to use optical systems; it can be shown that a coherent optical system, using a complex filter corresponding to the pattern being sought, can perform the required correlation operation. The method of setting up the coherent optical system and of constructing the complex filter are described. The limitations of the recording material used in making the filter are discussed and results are presented showing the effect of these limitations on the output of the system. Problems concerned with the practical use of the optical correlator are discussed, including the effect of optical aberrations and difficulties in using photographic input data. A method of overcoming some of the recording limitations is also presented. Finally, some applications are described in which complex spatial filters have been used. The merits and disadvantages in this approach to pattern recognition are discussed and some suggestions for future applications are given.