python - Wrong ifft in numpy

Question

I have this bidimensional array called a:

[2.1814 + 0.0000i   1.1375 + 0.0000i   0.2506 + 0.0000i   0.2506 + 0.0000i   1.1375 + 0.0000i]
[-2.1719 + 0.2033i  -1.1326 + 0.1060i  -0.2496 + 0.0234i  -0.2496 + 0.0234i  -1.1326 + 0.1060i]
[2.0884 + 0.6302i   1.0890 + 0.3286i   0.2400 + 0.0724i   0.2400 + 0.0724i   1.0890 + 0.3286i]
[1.8018 + 1.2295i   0.9396 + 0.6412i   0.2070 + 0.1413i   0.2070 + 0.1413i   0.9396 + 0.6412i]

When I run ifft(a) in matlab it gives me this answer:

[0.9749 + 0.5158i   0.5084 + 0.2690i   0.1120 + 0.0593i   0.1120 + 0.0593i   0.5084 + 0.2690i]
[0.2798 - 1.1510i   0.1459 - 0.6002i   0.0322 - 0.1323i   0.0322 - 0.1323i   0.1459 - 0.6002i]
[1.1599 - 0.2007i   0.6049 - 0.1046i   0.1333 - 0.0231i   0.1333 - 0.0231i   0.6049 - 0.1046i]
[-0.2333 + 0.8359i  -0.1217 + 0.4359i  -0.0268 + 0.0960i  -0.0268 + 0.0960i  -0.1217 + 0.4359i]

When I run np.fft.ifft2(a) in python it gives me this answer

[[ 0.44314676+0.23444054j  0.22157338+0.11722027j  0.04431468+0.02344405j 0.04431468+0.02344405j  0.22157338+0.11722027j]
 [ 0.12718937-0.52317671j  0.06359468-0.26158836j  0.01271894-0.05231767j 0.01271894-0.05231767j  0.06359468-0.26158836j]
 [ 0.52724743-0.09120678j  0.26362372-0.04560339j  0.05272474-0.00912068j 0.05272474-0.00912068j  0.26362372-0.04560339j]
 [-0.10604754+0.37994295j -0.05302377+0.18997148j -0.01060475+0.0379943j -0.01060475+0.0379943j  -0.05302377+0.18997148j]]

Why is it different? How can I make both python and matlab give me the same answer?

question from:https://stackoverflow.com/questions/65866773/wrong-ifft-in-numpy

Answer 1

It is because you should use ifft2 in MATLAB, or np.fft.ifft in numpy.

ifft2 is two-dimensional inverse discrete Fourier Transform, and ifft is one-dimensional. These are two different algorithms and they lead to different answers.