If the image rows and/or columns are not powers of two then the image size if filled out to the next factor of two and padded with zeroes.
If a box is specified, then the operations are only carried out over the box.
The transform is done in floating point, as a default, the modulus of the transform is written out. The keywords real and imag and power can be used to over-ride this selection, by writing the real, imaginary and total power values of the transform instead of the square root of the power. If the absolute value of the transform exceeds the range of an integer, the data are scaled to 32768. An optional scale factor can be entered to over-ride this selection.
The zero spatial frequency point is placed in the center of the transformed image.
Several messages are printed as the routine progresses if the verbose keyword is given.
The input file is assumed to be the real part. If you wish to do a complex transform the file containing the imaginary part can be specified by using the imag= option.