There's an opencv image which I split into 3 channels:
image #opencv image
img_red = image[:, :, 2]
img_green = image[:, :, 1]
img_blue = image[:, :, 0]
Then there are three filters:
red_filter
green_filter
blue_filter
Which are all numpy arrays, but are mostly populated by zeroes so that the format looks something like this:
[0, 0, 0, 132, ... 0, 15, 0, 230, 0]
...
[32, 0, 5, 0, ... 0, 2, 150, 0, 0]
I'd like to use every nonzero value in these filters to overwrite the same index in the channels.
Something like this:
img_red[index] = red_filter[index] if red_filter != 0
img_green[index] = green_filter[index] if green_filter != 0
img_blue[index] = blue_filter[index] if blue_filter != 0
final_img = cv2.merge(img_red, img_green, img_blue)
For example if the channel would look like this:
[44, 225, 43, ... 24, 76, 56]
And the filter:
[0, 0, 25 ... 2, 0, 91]
Then the result should be:
[44, 225, 25 ... 2, 76, 91]
I've tried using for loops and list comprehensions, but this code would have to be run over every frame in a video, so I'm wondering if there's a faster way to achieve the same result.
Is there some sort of image filtering in opencv, or numpy operation that would fulfill this process efficiently?
It seems like you're looking np.where method:
channel = np.array([44, 225, 43, 24, 76, 56])
filter = np.array([0, 0, 25, 2, 0, 91])
#result = np.array([44, 225, 25, 2, 76, 91])
>>> np.where(filter==0, channel, filter)
array([ 44, 225, 25, 2, 76, 91])