Replace numpy arrays of arrays with values

Question

I am trying to convert a numpy array of arrays. An example input is this:

np.array([[[0, 0, 0], [255, 255, 255]], [[255, 255, 255], [0, 0, 0]]], np.uint8)

I want to replace all arrays with values [0, 0, 0] with 0 and the [255, 255, 255] with 255. I want the final result to be in this format:

np.array([[0, 255], [255, 0]], np.uint8)

The array will always have the above values but will be different size.

This is the code I created for doing this:

array_list = []
for row in input_image:
    row_list = []
    for item in row:
        if np.array_equal(item, [0, 0, 0]):
            row_list.append(0)
        else:
            row_list.append(255)
    array_list.append(row_list)
output_image = np.array(array_list, np.uint8)

The above code is very slow for big arrays and I was thinking that there might be a way to do this directly with numpy, though I couldn't find a way. Do you have any suggestions for doing this more efficient?

What you show to a multi-dimensional array that you want to reduce along the third axis. This is a common operation. However, what do you want to do with other values such as [1, 128, 32], or even [255, 0, 0]? — Mad Physicist
– Mad Physicist, Commented Feb 8, 2020 at 14:33
I don't have any other values because I am fixing it with this: input_image = np.where(input_image != 255, 0, input_image) — bkbilly
– bkbilly, Commented Feb 8, 2020 at 14:35
That doesn't guarantee that you don't have say [255, 0, 255] as a value. How do your plan on handling that? — Mad Physicist
– Mad Physicist, Commented Feb 8, 2020 at 14:42
If a is the input, you could use a[:, :, 0], or a[..., 0]. — Warren Weckesser
– Warren Weckesser, Commented Feb 8, 2020 at 14:50
Yes, you are right, I didn't thought of that, though it won't be a problem on my application. All I want is to convert some black and white images to masks for OpenCV. — bkbilly
– bkbilly, Commented Feb 8, 2020 at 14:51

FBruzzesi · Accepted Answer · 2020-02-08 14:39:18Z

2

X_new = np.mean(X, axis=2)
X_new
array([[  0., 255.],
       [255.,   0.]])

edited Feb 8, 2020 at 14:39

answered Feb 8, 2020 at 14:29

FBruzzesi

6,6143 gold badges19 silver badges42 bronze badges

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

3 Comments

Mad Physicist Over a year ago

Just take the mean across the third axis. Don't loop with numpy if it's at all avoidable.

Mad Physicist Over a year ago

And it's always avoidable. The only time I haven't been a able to find a fully vectorized solution was when I wanted to solve multiple simultaneous least squares problems with different coefficients.

FBruzzesi Over a year ago

@bkbilly Consider to accept the answer as described here stackoverflow.com/help/someone-answers

Collectives™ on Stack Overflow

Replace numpy arrays of arrays with values

1 Answer 1

3 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

3 Comments

Your Answer

Sign up or log in

Post as a guest

Related