Thu, 22 Jul 2010 00:47:20 -0400, Robin Kraft wrote:
[clip]
> Let's say the image looks like this: np.random.randint(0,2,
> 16).reshape(4,4)
>
> array([[0, 0, 0, 1],
> [0, 0, 1, 1],
> [1, 1, 0, 0],
> [0, 0, 0, 0]])
>
> I want to use a square, non-overlapping moving window for resampling, so
> that I get a count of all the 1's in each 2x2 window.
>
> 0, 0, 0, 1
> 0, 0, 1, 1 0 3
> => 2 0
> 1, 1, 0, 0
> 0, 0, 0, 0
>
> In another situation with similar data I'll need the average, or the
> maximum value, etc..
Non-overlapping windows can be done by reshaping:
x = np.array([[0, 0, 0, 1, 1, 1],
[0, 0, 1, 1, 0, 0],
[1, 1, 0, 0, 1, 1],
[0, 0, 0, 0, 1, 1],
[1, 0, 1, 0, 1, 1],
[0, 0, 1, 0, 0, 0]])
y = x.reshape(3,2,3,2)
y2 = y.sum(axis=3).sum(axis=1)
# -> array([[0, 3, 2],
# [2, 0, 4],
# [1, 2, 2]])
y2 = x.reshape(3,2,3,2).transpose(0,2,1,3).reshape(3,3,4).sum(axis=-1)
# -> array([[0, 3, 2],
# [2, 0, 4],
# [1, 2, 2]])
The above requires no copying of data, and should be relatively fast. If
you need overlapping windows, those can be emulated with strides:
http://mentat.za.net/numpy/scipy2009/stefanv_numpy_advanced.pdf
http://conference.scipy.org/scipy2010/slides/tutorials
/stefan_vd_walt_numpy_advanced.pdf
--
Pauli Virtanen
_______________________________________________
NumPy-Discussion mailing list
NumPy-Discussion@scipy.org
http://mail.scipy.org/mailman/listinfo/numpy-discussion
