IDL rebin in python

rebin.py

import numpy as np

def rebin(a, new_shape):
    """
    Resizes a 2d array by averaging or repeating elements, 
    new dimensions must be integral factors of original dimensions

    Parameters
    ----------
    a : array_like
        Input array.
    new_shape : tuple of int
        Shape of the output array

    Returns
    -------
    rebinned_array : ndarray
        If the new shape is smaller of the input array, the data are averaged, 
        if the new shape is bigger array elements are repeated

    See Also
    --------
    resize : Return a new array with the specified shape.

    Examples
    --------
    >>> a = np.array([[0, 1], [2, 3]])
    >>> b = rebin(a, (4, 6)) #upsize
    >>> b
    array([[0, 0, 0, 1, 1, 1],
           [0, 0, 0, 1, 1, 1],
           [2, 2, 2, 3, 3, 3],
           [2, 2, 2, 3, 3, 3]])

    >>> c = rebin(b, (2, 3)) #downsize
    >>> c
    array([[ 0. ,  0.5,  1. ],
           [ 2. ,  2.5,  3. ]])

    """
    M, N = a.shape
    m, n = new_shape
    if m<M:
        return a.reshape((m,M/m,n,N/n)).mean(3).mean(1)
    else:
        return np.repeat(np.repeat(a, m/M, axis=0), n/N, axis=1)

if __name__ == '__main__':
    import doctest
    doctest.testmod()

Came across your method from stackoverflow and wanted to let you know it only works for regular numpy arrays. This method produces wrong results for masked arrays.
-marius

Hey I have written a version of this function for ndarrays with arbitrary dimensions:
https://gist.github.com/derricw/95eab740e1b08b78c03f
It also lets you sum or average.