Is there a numpy function to replace one array with the same part of another array

Question

I have two images stored as numpy arrays. I am looking for a function that can replace the parts of array2 that have black pixels ([0, 0, 0]) with the same indices of array1. My code so far is non-numpy:

for y in range(array2.shape[0]):
            for x in range(array2.shape[1]):
                if np.all(array2[y, x]) == False:
                    array2[y, x] = array1[y, x]

The code loops over every pixel in array2 to see if all of the channels are 0 (black), and if so copies the colour of array1 at that index onto array2.

This is obviously very slow because it loops over a lot of pixels that aren't black and has no affect on them. I assume numpy has a function that can do this, but I can't understand what they do. Any guidance (with or without numpy) is appreciated.

could you please add a minimal example for array2 and array1? at least to have matching .shape — Adam.Er8
– Adam.Er8, Commented Jul 9, 2020 at 12:27

avysk · Accepted Answer · 2020-07-09 14:03:18Z

2

Is this something like you want? (I assume that there are no negative values.) (Edit: but see comment by Adam.Er8 to lift this limitation)

>>> arr1                                                                                                                                             
array([[[1, 1, 1],
        [2, 2, 2],
        [3, 3, 3]],

       [[4, 4, 4],
        [5, 5, 5],
        [6, 6, 6]],

       [[7, 7, 7],
        [8, 8, 8],
        [9, 9, 9]]])
>>> arr2
array([[[1, 0, 0],
        [0, 1, 0],
        [0, 0, 1]],

       [[1, 0, 0],
        [0, 1, 0],
        [0, 0, 0]],

       [[1, 0, 0],
        [0, 0, 0],
        [0, 0, 0]]])
>>> idx = arr2.sum(axis=2) == 0
>>> idx
array([[False, False, False],
       [False, False,  True],
       [False,  True,  True]])
>>> arr2[idx] = arr1[idx]
>>> arr2
array([[[1, 0, 0],
        [0, 1, 0],
        [0, 0, 1]],

       [[1, 0, 0],
        [0, 1, 0],
        [6, 6, 6]],

       [[1, 0, 0],
        [8, 8, 8],
        [9, 9, 9]]])

edited Jul 9, 2020 at 14:03

answered Jul 9, 2020 at 11:43

avysk

2,04514 silver badges22 bronze badges

Sign up to request clarification or add additional context in comments.

8 Comments

Adam.Er8 Over a year ago

great answer! note: you can use idx = (arr2 == 0).all(axis=1) and not rely on sum and non-negative values.

incarnadine Over a year ago

Close, but I am looking for every zero in arr2 to be replaced with its value in arr1, not just the lines that are only zeros.

Adam.Er8 Over a year ago

@incarnadine you probably have a 3d array, which is a matrix of [R,G,B] pixels, which is exactly what's described in this answer.

avysk Over a year ago

@incarnadine that's what you've described in the question: you want black pixels (triples of 0) to be replaced, not every 0. For every 0 it's even simpler, just use idx = arr2 == 0.

Adam.Er8 Over a year ago

@avysk and it gave the same answer! lol! subtle bugs >_<

|

Paul Panzer · Accepted Answer · 2020-07-09 13:49:48Z

1

You can use np.copyto:

from scipy.misc import face
from PIL import Image
 
img = face()
out = np.random.randint(-1000,256,img.shape).clip(0,255).astype(np.uint8)

Image.fromarray(out).show()

np.copyto(out,img,where=(out==0).all(axis=-1,keepdims=True))

Image.fromarray(out).show()

answered Jul 9, 2020 at 13:49

Paul Panzer

53.3k3 gold badges59 silver badges103 bronze badges

Collectives™ on Stack Overflow

Is there a numpy function to replace one array with the same part of another array

2 Answers 2

8 Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

8 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related