Question

I am trying to reimplement in python an IDL function:

http://star.pst.qub.ac.uk/idl/REBIN.html

which downsizes by an integer factor a 2d array by averaging.

For example:

>>> a=np.arange(24).reshape((4,6))
>>> a
array([[ 0,  1,  2,  3,  4,  5],
       [ 6,  7,  8,  9, 10, 11],
       [12, 13, 14, 15, 16, 17],
       [18, 19, 20, 21, 22, 23]])

I would like to resize it to (2,3) by taking the mean of the relevant samples, the expected output would be:

>>> b = rebin(a, (2, 3))
>>> b
array([[  3.5,   5.5,  7.5],
       [ 15.5, 17.5,  19.5]])

i.e. b[0,0] = np.mean(a[:2,:2]), b[0,1] = np.mean(a[:2,2:4]) and so on.

I believe I should reshape to a 4 dimensional array and then take the mean on the correct slice, but could not figure out the algorithm. Would you have any hint?

Answer 1

Here’s an example based on the answer you’ve linked (for clarity):

>>> import numpy as np
>>> a = np.arange(24).reshape((4,6))
>>> a
array([[ 0,  1,  2,  3,  4,  5],
       [ 6,  7,  8,  9, 10, 11],
       [12, 13, 14, 15, 16, 17],
       [18, 19, 20, 21, 22, 23]])
>>> a.reshape((2,a.shape[0]//2,3,-1)).mean(axis=3).mean(1)
array([[  3.5,   5.5,   7.5],
       [ 15.5,  17.5,  19.5]])

As a function:

def rebin(a, shape):
    sh = shape[0],a.shape[0]//shape[0],shape[1],a.shape[1]//shape[1]
    return a.reshape(sh).mean(-1).mean(1)