Numpy Image Array Replace specific color with 1, other colors with 0

Question

I have to Encode an Image for Semantic Segmentation. The Input Image is of shape (128, 256, 3) with 128 x 256 RGB Values. I want an Output Shape of (128, 256) where every 1 represent, that the pixel matched the given Color and 0 represents, that another RGB Value was present.

[[20, 20, 20], [30, 30, 30], [40, 40, 40]] with the Filter [20,20,20] should result in [1, 0, 0]

Any Help would be greatly appreciated.

Optimally This Method should be feasible to apply to an array of shape (16, 128, 256, 3) with 16 pictures in it, applying the filter to every picture.

Correct me if I'm wrong, your example, the matrix [[20, 20, 20], [30, 30, 30], [40, 40, 40]] is 2D, so the filter should be 20 only and the result will be [[1,1,1],[0,0,0],[0,0,0]], is it correct? — CuCaRot
– CuCaRot, Commented Jun 17, 2021 at 8:13
The Matrix in the Question is just an example. The Natrix in my Code is of shape (16, 128, 256, 3) or (8, 128, 1256, 3) based on my batch_size. — josuaschenk
– josuaschenk, Commented Jun 17, 2021 at 8:18

Udi · Accepted Answer · 2021-06-17 09:05:08Z

1

Using np.where for one image:

filter_pixel = np.array([20, 20, 20])
image = np.array([[[20, 20, 20], [30, 30, 30], [40, 40, 40]],
                  [[20, 10, 20], [20, 20, 20], [40, 40, 40]]])

new_image = np.where(np.all(image == filter_pixel, axis=2), 1,0)
print(new_image)

Output:

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

edit, for a number of images:

filter_pixels = np.array([[20, 20, 20], [30, 30, 30]])
filter_pixels = filter_pixels[:, np.newaxis, np.newaxis, :]
images = np.array([[[[20, 20, 20], [30, 30, 30], [40, 40, 40]],
                    [[20, 10, 20], [20, 20, 20], [40, 40, 40]]],
                  [[[20, 20, 20], [30, 30, 30], [40, 40, 40]],
                   [[20, 10, 30], [20, 20, 20], [30, 30, 30]]]])

new_images = np.where(np.all(images == filter_pixels, axis=3), 1, 0)

print(new_images)

Output:

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

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

edited Jun 17, 2021 at 9:05

answered Jun 17, 2021 at 8:26

Udi

1721 silver badge6 bronze badges

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

7 Comments

josuaschenk Over a year ago

The Problem with this approach is that it returns 1 if one element matches the filter. e.g. image = np.array([[20, 20, 20], [20, 30, 30], [40, 40, 40]]) would result in [1, 1, 0] which is not what i want

Udi Over a year ago

Yea , you are right. Just edited to make it work.

josuaschenk Over a year ago

Thanks, I figured it out with your approach. I had to use axis=3 since my array has one dimension more than your test array, but it works just fine and is computationally fast. Thanks a Lot!

Udi Over a year ago

I added the complete answer for you question, it now works for a number of images.

josuaschenk Over a year ago

The Array I have after succesfully encoding it always consists of one one and the rest zeros. So for example the output could be [1,0,0,0,0,0,0,0,0,0,0,0]. This would be interpreted as class 0 (from 12). Now i want to replace these classes with the respective rgb values again, so if class 0 corrosponds to RBG [80, 80, 80] i want to insert that into the array changing its shape from (16, 128, 256, 12) to (16, 128, 256, 3) again. The output then should match the before existing array which I encoded

|

Collectives™ on Stack Overflow

Numpy Image Array Replace specific color with 1, other colors with 0

1 Answer 1

7 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

7 Comments

Your Answer

Sign up or log in

Post as a guest

Related