I'm trying to implement fft and gaussian low pass filter multiplication to 16x16 image.

As I know fft can be separable. What if there are 16x16 one image, then is this the same ? between 4x4 part of the whole 16x16 fft result and just only 4x4 fft result?

If they are same then I think that the latter method is more resource efficienciable.

enter image description here


1 Answer 1


I don't think "separable" means what you think it means – it means that you can implement the 2 DFT as the concatenation of a $N_{rows}$ amount of $N_{columns}$ long DFTs followed by $N_{columns}$ DFTs of length $N_{rows}$.

So, no, your approach is not correct.

  • $\begingroup$ I agree. so I modify to 'partial'. As i know H.264 and MPEG sort of things are using these methods to manipulate the image with 4x4 and 8x8 and 16x16 macro block with fft. so I thought that I can implement that kinds of way. $\endgroup$
    – start01
    Mar 11, 2018 at 12:32
  • $\begingroup$ No, you can't. Really, maybe sit down and look at the formula for the DFT. It's really apparent. $\endgroup$ Mar 11, 2018 at 12:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.