I think I understand the working of ndimage.generic_filter but I don't understand what happens when you throw in an array with more than 2 dimensions (like [[R],[G],[B]] values from an image)
Given the following code:
import numpy as np
from scipy.ndimage import generic_filter
a = np.arange(36).reshape(3,3,4)
def fnc(buffer):
return np.min(buffer)
footprint = [
[[1, 1], [1, 1]],
[[1, 1], [1, 1]],
[[1, 1], [1, 1]]
]
generic_filter(a, fnc, footprint = footprint)
The output is:
array([[
[ 0, 0, 1, 2],
[ 0, 0, 1, 2],
[ 4, 4, 5, 6]
],[
[ 0, 0, 1, 2],
[ 0, 0, 1, 2],
[ 4, 4, 5, 6]
],[
[12, 12, 13, 14],
[12, 12, 13, 14],
[16, 16, 17, 18]
]])
It seems to me here that the first array is duplicated at the cost of the last entered array which I can't see the logic of. Would this be intentional behaviour or am I not supposed to use generic_filter like this?
