[FIXED] Convert image to BGRA color space without using opencv

Issue

Is it possible to convert an RGBA image to BGRA without using the following code, i.e. without using opencv?

image = self.cv2.cvtColor(image, self.cv2.COLOR_RGBA2BGRA)

Solution

Like this with "fancy indexing":

# Make a dummy, random 4-channel image
RGBA = np.random.randint(0,256,(2,3,4), np.uint8)

In [3]: RGBA
Out[3]: 
 array([[[102, 204,  36, 128],
         [178, 151, 166,  45],
         [199,  49, 104,  98]],

        [[ 79,  33, 223,  62],
         [ 26,  34, 233, 254],
         [ 62,  20,  57, 149]]], dtype=uint8)

# Convert RGBA to BGRA
BGRA = RGBA[..., [2,1,0,3]]

In [5]: BGRA
Out[5]: 
array([[[ 36, 204, 102, 128],
        [166, 151, 178,  45],
        [104,  49, 199,  98]],

       [[223,  33,  79,  62],
        [233,  34,  26, 254],
        [ 57,  20,  62, 149]]], dtype=uint8)

I think OpenCV likes its data contiguous, so if you get issues, use:

BGRA = RGBA[..., [2,1,0,3]].copy()

Answered By - Mark Setchell