ABSTRACT
There has been considerable interest in using correlators for pattern recognition. Correlators are inherently shift invariant, allowing us to locate patterns (such as moving targets) in the input scene merely by locating the correlation peak. Thus, we do not need to segment or register the images prior to correlation, as we have to do in alternate methods for pattern recognition. Accordingly, a wide variety of correlation algorithms and architectures have been investigated (both optical and digital) to deal with the important problem of finding patterns in images.
