Correlation and template matching and matched filtering.

Imaging that the image is binary – glowing light and darkness. A binary template has the shape of the glowing light. When the template is perfectly aligned with the object, maximum light transmission occurs. Misalignment causes less light to be transmitted. Correlation and/or convolution allow us to picture all alignments at once. What has just been described is an autocorrelation that always peaks at the origin (perfect alignment). Cross correlation occurs when the template and object are not identical. The Fourier plane filter that produces autocorrelation is the matched filter. Here is more information on these topics.

http://bigwww.epfl.ch/demo/templatematching/correlation.html

http://www.khoral.com/contrib/contrib/dip2001/html-dip/c6/s5/front-page.html

There is a very useful advanced version called progressive correlation.

http://maya.ctr.columbia.edu:8080/Template_matching/progressive_template_matching.html