Abstract

In optical correlation processes, the intensities of correlation peaks generally serve as the tool for decision-making process in pattern recognition use. The distribution at the correlation plan provides the information for determining the existence of the object being recognized in the references and its whereabouts. However, in a binary pattern recognition process, this distribution at the correlation plane also provides information about the contour of the input object if references are properly chosen. This type of information offers a possibility of updating the reference image when a good match between the input object and the reference images does not exist. Consequently, the reference image can be modified toward the shape of the input image so that an adaptive recognition process can be achieved. In this paper, interpretation of the distribution of correlated result toward the contour of binary input image is investigated. A generalization of these interpretations is formulated and presented. An optical correlator based on joint transform architecture is used to verify the generalized interpretation and the proposed adaptive recognition process.

© 1991 Optical Society of America