Is there an easy elegant way to do a 3x3 convolution just by using numpy (no scipy.ndimage allowed)?
For example, suppose I have following 2-D array
array([[ 0, 3, 6, 9],
[12, 15, 18, 21],
[24, 27, 30, 33],
[36, 39, 42, 45]])
I want to produce following sum of neighbors using (3x3) convolution and 0 padding
array([[ 30, 54, 72, 54],
[ 81, 135, 162, 117],
[153, 243, 270, 189],
[126, 198, 216, 150]])
