Replace 3D array elements with numpy

Question

I have a 3D array filled with RGB values and I need to replace some "pixels" with another value as fast as possible.

The array looks like this:

[[[ 78  77  75]
  [ 72  70  67]
  [ 72  70  67]
  ...
  [ 73  74  73]
  [ 71  72  71]
  [ 66  67  67]]]

I used numpy select to replace some values but this is not what I need as it doesn't check for the whole pixel RGB.

np.select([array > 77, array < 77, [0, 0], array)

I tried to compare the whole RGB array with another one but it didn't work.

np.select([array != [72, 70, 77]], [[0, 0, 0]], array)

Creating a mask didn't help either. I wasn't able to replace the mask with an RGB array.

color = [30, 30, 57]
mask = np.any(array != color, axis=2)

Answer 1

You can define your two conditions, combine it using logical_or(), and then use where():

import numpy as np
 
array = np.array([[[78, 77, 75],
                  [72, 70, 67],
                  [72, 70, 67]],
                  [[78, 77, 75],
                  [72, 70, 67],
                  [72, 70, 67]],
                  [[73, 74, 73],
                  [71, 72, 71],
                  [66, 67, 67]]])
 
color = [30, 30, 57]
cond_a = array[:, :, 0] > 77
cond_b = array[:, :, 1] < 77
conds = np.logical_or(cond_a, cond_b)
res = np.where(conds, color, array)
 
print(res)

Prints

[[[30 30 57]
  [30 30 57]
  [30 30 57]]

 [[30 30 57]
  [30 30 57]
  [30 30 57]]

 [[30 30 57]
  [30 30 57]
  [30 30 57]]]