No products are associated with this question.
Basically you multiply the FFT (spectrum, Fourier Domain) by something that zeros out or attenuates higher frequencies. This "truncates" the spectrum. You choose the window size. It does not depend on the original image size, except that if the width of the truncating window is too huge (wider than the spectrum), you wouldn't be truncating anything. So in a very indirect way it depends on the image size (because hte spectrum size depends on the image size and thus the largest truncating window depends on the image size), but not really directly.
I used some thing like your question when i worked with Fourier transform. And the analysis could be on the frequecy domain (frequency spectrum)
Or a window of a cicle is used to generate the RMS value of a wave of many of cicles. step by step. (its a way of windowing).
I did not sure if it is your question.