python - Opencv or Numpy-- replace a list of pixels in an image, efficiently

Question

Hi OpenCV or Numpy Gurus,

I've been searching to an answer to this, but I'm surprised to not find it here or elsewhere...

I have a black image, and I want to replace a list of pixels (quite a large list) with a certain value. In the future, the "certain value" will be a list of values, but for the moment, let's keep it simple.

Simplifying even further, I just use a 2-d array of single numbers below, instead of RGB values... So, this fairly inefficient approach works for this sort of thing:

a = np.zeros((5, 5,1))

for i in np.asarray( ([2,3],[3,4]) ):
       a[i[0], i[1]] = 20

I was hoping there was a way to use np.put on a two dimensional array, where I don't rely on a python loop.

Now I had the suggestion that I could use

a[ ([2,3],[3,4]) = 20

However I notice when I use this approach in my real OpenCV problem it does not work. This very inefficient approach does:

coords_list = ([3,5],[55,60],[25,90])
black_image =np.zeros((480,640,3))
for i in coords_list:
    black_image[i[1],i[0]] = [255,255,255]

I expect this is a simple question to many out there.

Thanks!

Answer 1

In fact, you can directly use your tuple of coordinates to index your matrix: a simple a[([2, 3], [3, 4])] = 20 gives the exact same result as your for loop here.