Pad NumPy Array based on Condition

Question

Given is a inp_arr of shape (m, n) with 0 s and 255 s as values. I want to pad the 0 s with k zeros in a linear_ramp mode.

• The 0 s are not necessarily next to each other, so there could be two groups of 0 s.
• The output dimension can be greater than (m, n)

I already tried this: np.where(inp_arr== 0, np.pad(inp_arr, 2, pad_with), inp_arr) , giving me a ValueError that shapes cannot be broadcast together. pad_width is taken from here .

Example

import numpy as np

inp_arr = np.array([[255, 255, 255, 255, 255, 255, 255, 255],
 [255, 255, 255, 255, 255, 255, 255, 255],
 [255, 0, 0, 255, 255, 255, 255, 255],
 [255, 0, 0, 255, 255, 255, 255, 255],
 [255, 255, 255, 255, 255, 255, 255, 255],
 [255, 255, 255, 255, 255, 255, 255, 255],
 [255, 255, 255, 255, 0, 0, 255, 255],
 [255, 255, 255, 255, 0, 0, 255, 255],
 [255, 255, 255, 255, 255, 255, 255, 255]])

out_arr = np.array([[255, 255, 255, 255, 255, 255, 255, 255],
 [0, 0, 0, 0, 255, 255, 255, 255],
 [0, 0, 0, 0, 255, 255, 255, 255],
 [0, 0, 0, 0, 255, 255, 255, 255],
 [0, 0, 0, 0, 255, 255, 255, 255],
 [255, 255, 255, 0, 0, 0, 0, 255],
 [255, 255, 255, 0, 0, 0, 0, 255],
 [255, 255, 255, 0, 0, 0, 0, 255],
 [255, 255, 255, 0, 0, 0, 0, 255]])

Thanks for your suggestions: :)

Answer 1

That looks like an erosion :

>>> from scipy.ndimage import grey_erosion
>>> grey_erosion(inp_arr, size=(3,3))
array([[255, 255, 255, 255, 255, 255, 255, 255],
       [  0,   0,   0,   0, 255, 255, 255, 255],
       [  0,   0,   0,   0, 255, 255, 255, 255],
       [  0,   0,   0,   0, 255, 255, 255, 255],
       [  0,   0,   0,   0, 255, 255, 255, 255],
       [255, 255, 255,   0,   0,   0,   0, 255],
       [255, 255, 255,   0,   0,   0,   0, 255],
       [255, 255, 255,   0,   0,   0,   0, 255],
       [255, 255, 255,   0,   0,   0,   0, 255]])