How to get a padded slice of a multidimensional array?

Question

I am stuck on a little issue in the project I am currently working on.

Getting straight to the point, let's assume I have a 2-dimensional numpy.array - I will call it arr.

I need to slice arr, but this slice must contain some padding depending on the selected interval.

Example:

arr = numpy.array([
    [  1,  2,  3,  4,  5],
    [  6,  7,  8,  9, 10],
    [ 11, 12, 13, 14, 15],
    [ 16, 17, 18, 19, 20],
    [ 21, 22, 23, 24, 25]
])

Actually, numpy's response for arr[3:7, 3:7] is:

array([[19, 20],
       [24, 25]])

But I need it to be padded as if arr were bigger than it really is.

Here is what I need as response for arr[3:7, 3:7]:

array([[19, 20,  0,  0],
       [24, 25,  0,  0],
       [ 0,  0,  0,  0],
       [ 0,  0,  0,  0]])

This padding should also occur in case of negative indices. If the requested slice is bigger than the whole image, padding must occur in all sides, if needed.

Another example, negative indices. This is the expected result for arr[-2:2, -1:3]:

array([[ 0,  0,  0,  0],
       [ 0,  0,  1,  2],
       [ 0,  0,  6,  7],
       [ 0,  0, 11, 12]])

Is there any native numpy function for this? If not, any idea of how can I implement this?

Answer 1

About the first part of your question you can use a simple indexing, and you can create a zero_like of your array with numpy.zeros_like then assign the special part :

>>> new=numpy.zeros_like(arr)
>>> part=arr[3:7, 3:7]
>>> i,j=part.shape
>>> new[:i,:j]=part
>>> new
array([[19, 20,  0,  0,  0],
       [24, 25,  0,  0,  0],
       [ 0,  0,  0,  0,  0],
       [ 0,  0,  0,  0,  0],
       [ 0,  0,  0,  0,  0]])

But for the second case you can not use a negative indexing for for numpy arrays like this.Negative indices are interpreted as counting from the end of the array so if you are counting from -2 actually in a 5x5 array there are not any row between -2 and 2 so the result would be an empty array :

>>> arr[-2:2]
array([], shape=(0, 5), dtype=int64)

Answer 2

You can do something like:

print np.lib.pad(arr[3:7,3:7], ((0, 2), (0, 2)), 'constant', constant_values=(0,0 ))



[[19 20  0  0]
 [24 25  0  0]
 [ 0  0  0  0]
 [ 0  0  0  0]]

For the negative indexing:

print np.lib.pad(arr[ max(0,-1):3 , 0:2 ], ((1, 0), (2, 0)), 'constant', constant_values=(0,0 ))

[[ 0  0  0  0]
 [ 0  0  1  2]
 [ 0  0  6  7]
 [ 0  0 11 12]]

Check here for reference