get_matrix_dft - Computes the discrete Fourier transform (DFT) of a matrix
Example compile flags (system dependent):
-DLINUX_X86_64 -DLINUX_X86_64_OPTERON -DGNU_COMPILER
-lKJB -lfftw3 -lgsl -lgslcblas -ljpeg -lSVM -lstdc++ -lpthread -lSLATEC -lg2c -lacml -lacml_mv -lblas -lg2c -lncursesw
const Matrix *input_re_mp,
const Matrix *input_im_mp
This routine computes the discrete Fourier transform (DFT) of a paire of
matrices representing a complex matrix. The matrix representing the imaginary
part can be NULL.
The first two argument are pointers to matrices for the real and imaginary
parts of the result. If they are null, then matrices of the appropriate sizes
are created. If they are the wrong size, they are resized. Finally, if they
are the right size, then the storage is recycled, as is.
NO_ERROR on success, and ERROR on failure, with an error message being
This routine simply wraps the appropriate fftw library functions for
convenient use with our our data structures. The fftw library can also be
used directly in conjunction with the kjb_library by including fftw3.h
(the makefile builder should know where to find everthing). This can be
preferable in some cases where one wants to exploit the tuning mechanism
from that library, and cut down on overhead. However, for ocasional use,
the savings won't be very great.
The default behaviour of this routine is to inherits fftw's normalization
strategy which is to return unnormalized transforms. This means that
taking the transform, and then inverting, requires division by
N=num_cols*num_rows. This behaviour can be changed using the
This software is not adequatedly tested. It is recomended that
results are checked independantly where appropriate.