PIL Image Convert from RGB to YCbCr Results in 4 Channels Instead of 3 and Behaves Like RGB

Question

Well, the title is pretty self explanatory. I have an image file that I want to separate into Y, Cb and Cr respectively. After opening the file, convert it from RGB (which is the default mode when opening an image file) into YCbCr and then turn it into an array using numpy.array(), it resulted in a 2D array with 4 channels, which is not I expected as according to the documentation in http://www.nmt.edu/tcc/help/pubs/pil/pil.pdf

Here is what I do in the interpreter:

ImageFile = Image.open('filePath', 'r')
ImageFile = ImageFile.convert('YCbCr')
ImageFileYCbCr = numpy.array(ImageFile)
ImageFileYCbCr

resulted in

array([[[103, 140, 133,  95],
    [140, 133,  91, 141],
    [132,  88, 141, 131],
    ..., 
    [129,  65, 146, 129],
    [ 64, 146, 130,  65],
    [146, 129,  64, 147]],

   [[129,  64, 147, 129],
    [ 62, 149, 130,  62],
    [149, 130,  62, 149],
    ..., 

And when I separate it into its channels

ImageFileY = copy.deepcopy(ImageFileYCbCr) # to make a separate copy as array is immutable
ImageFileY[:,:,1] *= 0
ImageFileY[:,:,2] *= 0
ImageFileY[:,:,3] *= 0
ImageFileYOnly = Image.fromarray(ImageFileY)
ImageFileYOnly.show()

It resulted in a red color channel as if it is an RGB. Ho can I get the Y, Cb, Cr values respectively?

EDIT: Numpy version 1.3, Python 2.6 Linux Backtrack 5

Answer 1

https://mail.python.org/pipermail/image-sig/2010-October/006526.html

It's an old bug with Numpy. To correct it

>>> import numpy
>>> import Image as im
>>> image = im.open('bush640x360.png')
>>> ycbcr = image.convert('YCbCr')

>>> B = numpy.ndarray((image.size[1], image.size[0], 3), 'u1', ycbcr.tostring())
>>> print B.shape
(360, 640, 3)
>>> im.fromarray(B[:,:,0], "L").show()

Answer 2

FYI, for future people coming from Google:

This seems to work now.

For reference, I have Pillow 6.1.0 and numpy 1.17.0. Doing

img = np.array(Image.open(p).convert('YCbCr'))

gives the same as

img = Image.open(p).convert('YCbCr')
img = np.ndarray((img.size[1], img.size[0], 3), 'u1', img.tobytes())

and is not the same as RGB.