如何用从另一个数组中采样的数字填充数组的切片？

Question

I have a (very) big grayscale image and a mask, both ranging from 0 to 1, stored as numpy arrays. The mask is full of 0s and 1s. I would like to turn all 1 pixels in the mask to resemble some noise from the image, which was already sampled to vector p .

I've tried mask[mask == 1] = np.random.choice(p) , but of course it fills the whole area with only one of the values in p . I would like to avoid a for loop, and I would like to avoid creating a new image filled entirely from p , as they are really big for my tiny laptop.

Any ideas?

Answer 1

Try to generate the random choice with size :

mask[mask==1] = np.random.choice(p, size=mask.sum())