FFT3D related question

[FranceBB]

Hi!
Some filters like FFT3D use the fast Fourier transform to get frequencies out of an image and get rid of the ones identified as noise.
My question is: why don't they use the Fast Discrete Cosine Transform instead?
FDCT is periodic in 2π instead of π, it has less discontinuity points and it's not in the complex domain (imaginary numbers).
So... why don't they use the FDCT instead?
Would be useful to make something like FDCT3D?

(It's 2.13 AM, I can't sleep and I started thinking about this xD)

[Chikuzen]

Hi!
Some filters like FFT3D use the fast Fourier transform to get frequencies out of an image and get rid of the ones identified as noise.
My question is: why don't they use the Fast Discrete Cosine Transform instead?
Probabry, libfftw was there.
If there was "Fastest Discrete Cosine Transform in East" project, maybe someone challenged that.
Would be useful to make something like FDCT3D?
I know there is a DCT based denoiser implementation.
https://github.com/norishigefukushima/RandomizedRedundantDCTDenoising
Is it useful? I don't know.
It's your business!

[feisty2]

Probabry, libfftw was there.
If there was "Fastest Discrete Cosine Transform in East" project, maybe someone challenged that.

http://www.fftw.org/doc/Real-even_002fodd-DFTs-_0028cosine_002fsine-transforms_0029.html

These are also known as types I-IV of the DCT and DST, and all four types are supported by FFTW's r2r interface

fftw does the dct/dst business too.