The article asks: Which is the better trade off? More color range per pixel, or more pixels with color channels?
The answer is more color range per pixel.
Chroma is subsampled compared to Luma because that's literally how vision works: the Human eye has far more Rods (brightness sensors) than Cones (color sensors) per unit area. Increasing the color resolution on a TV[1] to match the luma resolution is literally a waste of bits[2]. On the other hand, the color receptors are very sensitive to the exact shade of color[3], so increasing color bit depth is not a waste.
[1] Increased color resolution on a computer monitor is useful because you sometimes closely examine a small section of the monitor -- and possibly lean in to examine it better.
[2] The old NTSC analog system used the same principle: color was encoded using less bandwidth than the luma.
[3] As the author mentions, this is best observed in a gradient where the TV must present the gradations in sufficiently small steps to fool the eye into seeing a continuous gradation.
Actually that's not how rods and cones work. In daylight or bright indoor light settings, the rods deactivate. That is, they generate no signal. Not to mention you've got the relationship backwards- there's far more cones than rods per unit area. This is why you can see with much higher detail in daylight settings, but in low light settings you see in very low resolution in black and white (with a blue tone since the rods are mostly only sensitive to blue/green light)
These are important facts for deciding, for instance, that alarm clocks, and shipboard navigation dashes should use only red light- since red light doesn't kill the sensitivity of your rods. It doesn't deactivate your night vision.
That's not to say you're not partly right, it's just the color/luminance processing doesn't happen in the retina. It happens in the visual cortex, or possibly with some help from processing that happens in or near the optic nerve.
If screens move to deep color, does that mean their contrast ratio would also change to show such great variations? Ie an extreme would be the brightness of the sun down to pure darkness. Or would it basically be the same as now, just we could fit more variations of color into the same contrast? It's hard to visualize because 8bpp color already seems like it covers every RGB I would need to see...
The ability of a screen to display large contrasts is important, and tends to improve with every generation of displays, but it's entirely separate from the way colors are represented digitally.
Doesn't decorrelating the quantization noise (dither) along with appropriate noise shaping make these two equivalent? It's been a long time since I've thought about this in context of video, but it should be possible to directly trade one for the other in the wire format, as far as I remember.
The answer is more color range per pixel.
Chroma is subsampled compared to Luma because that's literally how vision works: the Human eye has far more Rods (brightness sensors) than Cones (color sensors) per unit area. Increasing the color resolution on a TV[1] to match the luma resolution is literally a waste of bits[2]. On the other hand, the color receptors are very sensitive to the exact shade of color[3], so increasing color bit depth is not a waste.
[1] Increased color resolution on a computer monitor is useful because you sometimes closely examine a small section of the monitor -- and possibly lean in to examine it better.
[2] The old NTSC analog system used the same principle: color was encoded using less bandwidth than the luma.
[3] As the author mentions, this is best observed in a gradient where the TV must present the gradations in sufficiently small steps to fool the eye into seeing a continuous gradation.