FFTW claims to be the fastest Fourier Transform in the West.
This is a high-level interface to libfftw.
We re-use the type classes from netlib-ffi.
Features:
Support of one (audio), two (image), three (video) dimensional data,
as well as n-dimensional one.
Support for batched transformations e.g. for block-wise Fourier transform.
No normalization like in plain fftw.
This is because I see no universally best place to apply normalization.
The use of comfort-array enables to use array shapes tailored
to Fourier transforms:
Shape.Cyclic- for data where the
n-1-th element
can also be accessed via index -1. Shape.Half- for complex data of a real-to-complex Fourier transform.
It saves you from case distinction between even and odd data length
and according mistakes.
Shape.Symmetric- for real-to-real Sine and Cosine transforms.
They assert that you will always use
the appropriate kind for back transformation.
For rather simple examples
see the packages align-audio and morbus-meniere.
See also package fft.
