Abstract

Pattern recognition requires the identification of an object with unknown orientation and variable scale. To solve this problem, multireference images can be used. However, there is a limit space-bandwidth product available in the optical system. A large number of references would essentially saturate the image space which would degrade the accuracy and the reliability of detection. To implement multiple references in the same plane, the information content of each reference image must be reduced. This reduction is achieved by using only the edge information in the reference image. In general, using an edge-enhanced image provides higher accuracy of detection, thus reducing the ambiguity of the spatial content. We note that the joint transform correlator has advantages such as simplicity in reference generation and lower resolution requirements, which are very suitable for electrically addressed low-resolution spatial light modulators such as LCTVs.

© 1988 Optical Society of America