Pattern Recognition Based On The Triple Correlation

In ordinary pattern recognition systems the input image is sent through a linear filter, such as a matched filter. The presence of the search pattern is identified as a peak in the output image. In some of the newer pattern recognition systems it is not the input image itself which is filtered. Instead, a "secondary signal" is tested by a suitable filtering process. A typical "secondary signal" is the power spectrum of the original input image. The power spectrum has lost the information of the location of the search pattern within the input. This loss of location information makes the job of signal processing easier, since the irrelevant data has been supressed. Hence, working with the power spectrum as a secondary signal is advantageous, provided the location of the search pattern is of no interest. Only the presence or absence of the search pattern has to be decided upon.