Name
fft - fft 2d on an image
Synopsis
fft [options] <in> <out>
Description
fft computes a Fast Fourier Transform on an input image. fft can also
compute an inverse FFT. Results can have swapped quadrants, and can be
expressed in polar or cartesian coordinates.
Take care about the formats : polar/cartesian and swapped/unswapped.
The default procedure is:
In input of a forward FFT, an image is required. It is taken as the
real part (in cartesian coordinates) of a complex 2d signal which imag-
inary part is set to zeros. The output of a forward FFT is a complex 2d
signal, i.e. 2 image planes. By default, the output is converted to
polar coordinates (modulus, phase), and then quadrants are swapped in
both modulus and phase to put low frequencies at the center of the
images. To prevent this default behaviour, use the -n and -c options
described below.
In input of an inverse FFT, a cube containing 2 images is required. It
is taken as a complex 2d signal which first plane contains the modulus
and second plane contains the phase. Before the inverse FFT is com-
puted, a swapping of quadrants occurs, and then a conversion to carte-
sian (real, imaginary) coordinates is performed. To prevent quadrant
swapping or cartesian conversion, use the -n and -c options described
below.
The output of an inverse FFT is a cube containing 2 images in cartesian
unswapped format, first one being the real part, second one being the
imaginary part (meaningless imagewise).
Algorithm
fft uses the Danielson-Lanczos lemma, in a code based on one originally
written by N. M. Brenner, described in Numerical Recipes in C.
Options
-c or --noconv
Switches to cartesian mode the following data:
output for a forward FFT (default is polar), or input for an
inverse FFT.
-i or --inverse
Inverse FFT. The input is a cube containing two planes only.
Default I/O format for these 2 planes is polar coordinates (mod-
ulus, phase). It will output a cube containing 1 plane only
(imaginary part is meaningless imagewise). The input will be
swapped before FFT, unless the -n option is used.
-n or --noswap
This option prevents fft from swapping the output of a forward
FFT and the input of an inverse FFT.
-s or --swaponly
This option can be used to apply the swapping and quit.
Swapping is done according to the following rule :
1 2
4 3
becomes then
3 4
2 1
which puts then the lowest frequencies at the center of the image.
Files
Input files shall all comply with FITS format.