Abstract

We propose an optical adaptive processing for pattern recognition by using a two-correlator system, which achieves an intensity invariant recognition by discriminating true correlation peaks from false correlation peaks with a dynamic threshold level proportional to the local intensity of input image. This processing is useful for detecting multiple objects having different intensities each other simultaneously, while a conventional processing can not always discriminate all objects from background because of applying a constant threshold level all over the image.

© 1995 IEEE