PIL fromarray for single channel image

Question

I am trying to obtain an image shaped (1080, 1920, 1) from one shaped (1080, 1920, 3) This is what I have been trying without success:

    for fr in fr_lst:
            frame = cv2.imread(os.path.join(frame_root, fr))

            #SPLIT CHANNELS
            frame = (cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
            r, g, b = cv2.split(frame)
            r = np.expand_dims(r, axis=2)
            print(r.shape)
            frame = Image.fromarray(r)

when I print the shape of r I get (1080, 1920, 1) but Image.fromarray(r) returns the error

TypeError: Cannot handle this data type: (1, 1, 1), |u1

I tried not expanding the dimensions, obtaining the shape of r of (1080,1920) and successfully running Image.fromarray(r)

I also tried to expand the dimensions of the PIL image frame = np.expand_dims(frame, axis=(2)) which seems to return the appropriate result, but has a strange behaviour:

If I use an array of size (1080, 1920, 3) and run size = frames[0].size I obtain size = 1920, 1080 which is great. But if I run size = frames[0].size with frames of shape (1080, 1920, 1) I obtain size = 2073600

My goal is to have an array of size (1920, 1080) when passing a frame of shape (1080, 1920, 1).

What am I doing wrong or not understanding?

Thank you

Answer 1

This doesn't make sense. If you have a BGR frame with shape (1920,1080,3) and you want a single channel image, it will have shape (1920,1080).

You can either take one channel:

singleChannel = im[..., 0]   # take Blue channel
singleChannel = im[..., 1]   # take Green channel
singleChannel = im[..., 2]   # take Red channel

Or you can convert to greyscale and take that:

singleChannel = cv.cvtColor(im, cv.COLOR_BGR2GRAY)