Black and white television came first. A black and white image only contains white, black and shades of grey. All these values could (and still can) be obtained merely by varying brightness.
Pure unadulterated sunlight is white (or at least white enough for our purposes). When light isn’t reflected (a black body), what we get is black. By varying the intensity of light, we get various shades of grey. Remember, grey is a perception of the human brain.
In scientific terms, the brightness of light is measured in terms of Luminance. The word ‘Luminance’ is directly defined by CIE in relation to human vision. It is represented by the letter Y as far as video is concerned.
As we have seen in the chapter on Gamma, brightness is ‘skewed’ for display, a process we call gamma compression. Instead of calling it Twisted Luminance, they decided to call video-encoded luminance ‘Luma’, represented by Y’.
Whenever we speak of the luminance values with respect to video, we are speaking of Luma (Y’). When a video signal is coming out of an SDI port as 4:2:2 or whatever, the correct notation for it is Y’CbCr. Notice the Y with an apostrophe? The Luminance is already ‘gammafied’, or has been applied with an encoding gamma.
From this chapter on RGB we know that if Y is the child of RGB, Y’ comes from R’G’B’.
Chrominance is the color information in a signal. When color television arrived, instead of starting from scratch they found a way to tack it on to Luma, as we have seen in the difference between CIE XYZ and xyY. From it, we know the two representations of color are named Cb and Cr.
It doesn’t mean green is neglected, mind you. From RGB you get XYZ, and from XYZ you derive xyY, which leads to Y’CbCr. This is what it looks like:
The top-most is the full RGB image. If it had to be split into Y’CbCr, Y’ would be the second, Cb the third and Cr the fourth. Don’t assume just because it’s a ‘b’ in Cb and an ‘r’ in Cr it means blue or red – just as if you don’t expect a man called Mr. Red to be red in color.
Chroma sampling and sub-sampling
The name of the game is ‘reducing bandwidth’. Luminance becomes Luma, the bit depth is reduced to 8, the color spaces (Rec. 709, PAL, NTSC) are reduced to half or less than half the range of the human eye, data is compressed using compression algorithms, and so on. Can we compress it even further?
Sure we could. Similar to how rods and cones work in the brain, they realized they could split up colors into Luminance and Chrominance – that’s right, our eye exhibits both RGB and Y’CbCr behavior. That’s why it works!
By a convoluted method of trial and error, and lots of testing, they discovered that they could actually throw out half or even three fourths of the color information and most people wouldn’t know it. People consume artificial flavoring every day, too.
It doesn’t work in the way data is compressed, and to really understand how it works, you’ll need to learn what sampling is. When only chrominance information (color) is sampled from analog data (or digital data), we call it Chroma Sampling.
When it is not sampled fully (like potato chips that don’t fill up the whole bag) the process is called Chroma sub-sampling.
It is a type of compression that can be applied at the time of recording in camera, or later in processing. Sub-sampling is what happens when you don’t sample 1:1, which means you throw out data. That’s the objective, to reduce bandwidth.
The most widely used model today is the Y’CbCr model. Take a look at this image:
From the above image you can see that all the four color boxes look the same, even though the left-most has much lesser color information than the right-most.
Is Y’CbCr a color space?
Y’CbCr is not a color space, nor is it strictly a color model. It is only an encoding system. You’ll find many folks confusing all these terms and ask questions like “Which color space should I use, YUV, 444 or RGB”? That’s like asking which cow should I be feeding, the one that looks like a cow, the one that looks like a cow with horns, or the same cow that I’m looking at from another angle?
Crazy? But that’s what happens when you don’t bother understanding your terms.
For high definition television, this is how it works out:
- RGB is the color model
- Rec. 709 is the color space
- Y’CbCr is the luma and chroma encoding system
- The numbers in Y’CbCr tells you the kind of luma and chroma sampling used
- 4:2:0 and 4:2:2 both have equal luma values, but dissimilar chroma values
- 3:1:1 and 4:2:2 have dissimilar luma and chroma values
Don’t forget – color model, color space, color encoding – three totally different things.