FFT for ndarrays
Switch branches/tags
Nothing to show
Latest commit ddfc848 Aug 11, 2016 @jaspervdg jaspervdg Tested non-standard strides.
It should make no difference (and it doesn't), but to be sure I added some padding to the 2D test cases.



Build Status

A fast Fourier transform implementation for ndarrays. You can use this to do image processing operations on big, higher dimensional typed arrays in JavaScript.


var zeros = require("zeros")
var ops = require("ndarray-ops")
var fft = require("ndarray-fft")

var x = ops.random(zeros([256, 256]))
  , y = ops.random(zeros([256, 256]))

//Forward transform x/y
fft(1, x, y)

//Invert transform
fft(-1, x, y)


Via npm:

npm install ndarray-fft

require("ndarray-fft")(dir, x, y)

Executes a fast Fourier transform on the complex valued array x/y.

  • dir - Either +/- 1. Determines whether to use a forward or inverse FFT
  • x the real part of the signal, encoded as an ndarray
  • y the imaginary part of the signal, encoded as an ndarray

x and y are transformed in place.

Note This code is fastest when the components of the shapes arrays are all powers of two. For non-power of two shapes, Bluestein's fft is used which is somewhat slower.

Note2 The inverse FFT is scaled by 1/N, forward FFT is unnormalized.


(c) 2013 Mikola Lysenko. MIT License.

Radix 2 FFT based on code by Paul Bourke.