Watch this Author's files
I found the answer of my previous question:
"The second term compensates for the dc-value of the kernel, to avoid unwanted dependence of the filter response on the absolute intensity of the image."
From "Distortion Invariant Object Recognition in the Dynamic Link Architecture"
Thanks for this submission.
I have a question about the equation used: why did you subtract 'exp ( -0.5 * Delt2 )' to the modulation component? I have never seen it before.