Correlation Filters for Optical Pattern Recognition

Abstract

Optical Pattern Recognition got its start when Vander Lugt proposed the frequency plane correlator to implement the matched spatial filtering. Since that time, correlators have been the subject of much research. While matched filter is theoretically optimum when the input signal (or image) is an exact replica of the reference signal, its performance degrades severely in practical situations where the input image may differ from the original due to distortions such as rotations and scale changes. Additional problems facing conventional correlators include: low light throughputs, lack of real time spatial light modulators and sensitivity to misalignment. Here, we review some of the significant developments in the area of optical correlation filter design that were aimed at solving some of these problems.

Indexing terms:

• Optical pattern recognition
• Optical correlation filter
• Spectral iteration