5 ms·

Fun fact: In human testing, subjects cannot tell the difference between 8-bit RGB and 10-bit RGB except for greyscale gradients. (I wish I could link the study

Fun fact: In human testing, subjects cannot tell the difference between 8-bit RGB and 10-bit RGB except for greyscale gradients. (I wish I could link the study, but I found it during the peak of my 2012-era graphics career. Suffice to say, more devs should study the science of perceptual testing.)

pixel_fcker 6y ago
For low-dynamic range maybe, but for HDR 8-bit is definitely not enough.
- sillysaurusx 6y ago
  If you measure this in a rigorous, scientific way, you will be forced to conclude that 8-bit RGB is enough. The reasons why are interesting, but it has to do with how the eye processes images and color. Remember, you can turn your screen as bright as you want. But it doesn't help the content appear more "real" or "better" or "HDR-like". Therefore, HDR is simply "the ability to discern between different levels of color." And if you try to measure exactly how many levels of color a human can discern, the answer is no more than the 256^3 available via 8-bit RGB. The precise claim: if you have two screens next to each other, one 8-bit RGB and one 10-bit RGB, and both screens are color calibrated / brightness calibrated identically, and the viewing conditions are identical for both screens, then if you blindfold someone and put them in front of a random screen, they will be unable to say "Ah, this is the 10-bit RGB screen!" more than random chance.
  - tjoff 6y ago
    If you show a grayscale gradient anyone not legally blind would pick the 10bit display in a fraction of a second.
    - sillysaurusx 6y ago
      Grayscale gradients, yes. Anything else, no. That was what researchers discovered.
      - tjoff 6y ago
        So the answer is: yes. 10-bit also comes with much better calibration precision where the translation isn't done in the display itself.
      - mixmastamyk 6y ago
        Monochrome, not grayscale. For example, many shades of blue in a high-contrast photo of the sky.
  - pixel_fcker 6y ago
    “If you calibrate the HDR screen to be as dim as the LDR screen they’ll look the same” isn’t a great argument.
    - sillysaurusx 6y ago
      So turn up the brightness of the LDR screen to match the HDR screen. The brightness isn't why HDR is HDR. I didn't say to turn down the brightness of the HDR screen. I said the brightness per area must match. Using maximum brightness often harms color reproducibility too, by the way. So if you believe brighter = better, don't.
      - pixel_fcker 6y ago
        No the dynamic range and wider color gamut is. Your argument is that 10-bit sRGB doesn’t give you anything over 8-bit sRGB, which I don’t agree with but it’s reasonable. But that’s not what HDR is. 8bpc is definitely not enough to represent HDR Rec.2020 or DCI-P3 without significant banding.
  - Jonnax 6y ago
    You're referencing a study, perhaps find it to back up your point. Because 8 vs 10 bit colour in sky photography is quite obvious
LockAndLol 6y ago
That wouldn't surprise me. If we couldn't tell them apart. There must also be limits to perception or resolution. Are our eyes even able to tell the difference between FHD, 2K and 4K on a 15" screen? I tested FHD vs 4K on a 31" screen and could not are a difference when watching a 10 minute video (I had 2 screens side by side for reference). On a semi related note. I also tried watching 60fps videos at 60Hz vs 144Hz (again side by side) to test out my new screen. No difference. Even fast moving things. I really feel like either I just don't perceive these things and it'll keep more money in my wallet, or that I'm not alone and everybody can save money. A chart of when differences actually become perceptible would be great. It wouldn't surprise me I'd the threshold for 4K were 42", 10bit only in grays and fps at a certain pixel per second movement speed.
eqvinox 6y ago
I can see the difference in my 8-bit desktop background gradient, which is blue, and I don't think that's surprising at all since it's a brightness/value difference which is probably perceived through more cells (or more sensitively) in the eye. I can't link a study either but I'd like to have 10-bit even if it's just for the light axis. Coincidentally in a YUV encoding I'd only push the V axis to 10 bit - maybe even 12 to cover extra HDR range, i.e. higher absolute brightness when the display supports it. (Or go with a logarithmic encoding, in which case even 8 bit might be enough...)
TwoBit 6y ago
Then they ran the test wrong. It's easy to construct 10 bit color images which are clearly discernable from 8 bit.

pixel_fcker 6y ago

For low-dynamic range maybe, but for HDR 8-bit is definitely not enough.

sillysaurusx 6y ago
If you measure this in a rigorous, scientific way, you will be forced to conclude that 8-bit RGB is enough. The reasons why are interesting, but it has to do with how the eye processes images and color. Remember, you can turn your screen as bright as you want. But it doesn't help the content appear more "real" or "better" or "HDR-like". Therefore, HDR is simply "the ability to discern between different levels of color." And if you try to measure exactly how many levels of color a human can discern, the answer is no more than the 256^3 available via 8-bit RGB. The precise claim: if you have two screens next to each other, one 8-bit RGB and one 10-bit RGB, and both screens are color calibrated / brightness calibrated identically, and the viewing conditions are identical for both screens, then if you blindfold someone and put them in front of a random screen, they will be unable to say "Ah, this is the 10-bit RGB screen!" more than random chance.
- tjoff 6y ago
  If you show a grayscale gradient anyone not legally blind would pick the 10bit display in a fraction of a second.
  - sillysaurusx 6y ago
    Grayscale gradients, yes. Anything else, no. That was what researchers discovered.
    - tjoff 6y ago
      So the answer is: yes. 10-bit also comes with much better calibration precision where the translation isn't done in the display itself.
    - mixmastamyk 6y ago
      Monochrome, not grayscale. For example, many shades of blue in a high-contrast photo of the sky.
- pixel_fcker 6y ago
  “If you calibrate the HDR screen to be as dim as the LDR screen they’ll look the same” isn’t a great argument.
  - sillysaurusx 6y ago
    So turn up the brightness of the LDR screen to match the HDR screen. The brightness isn't why HDR is HDR. I didn't say to turn down the brightness of the HDR screen. I said the brightness per area must match. Using maximum brightness often harms color reproducibility too, by the way. So if you believe brighter = better, don't.
    - pixel_fcker 6y ago
      No the dynamic range and wider color gamut is. Your argument is that 10-bit sRGB doesn’t give you anything over 8-bit sRGB, which I don’t agree with but it’s reasonable. But that’s not what HDR is. 8bpc is definitely not enough to represent HDR Rec.2020 or DCI-P3 without significant banding.
- Jonnax 6y ago
  You're referencing a study, perhaps find it to back up your point. Because 8 vs 10 bit colour in sky photography is quite obvious

LockAndLol 6y ago

That wouldn't surprise me. If we couldn't tell them apart. There must also be limits to perception or resolution. Are our eyes even able to tell the difference between FHD, 2K and 4K on a 15" screen? I tested FHD vs 4K on a 31" screen and could not are a difference when watching a 10 minute video (I had 2 screens side by side for reference). On a semi related note. I also tried watching 60fps videos at 60Hz vs 144Hz (again side by side) to test out my new screen. No difference. Even fast moving things. I really feel like either I just don't perceive these things and it'll keep more money in my wallet, or that I'm not alone and everybody can save money. A chart of when differences actually become perceptible would be great. It wouldn't surprise me I'd the threshold for 4K were 42", 10bit only in grays and fps at a certain pixel per second movement speed.

eqvinox 6y ago

I can see the difference in my 8-bit desktop background gradient, which is blue, and I don't think that's surprising at all since it's a brightness/value difference which is probably perceived through more cells (or more sensitively) in the eye. I can't link a study either but I'd like to have 10-bit even if it's just for the light axis. Coincidentally in a YUV encoding I'd only push the V axis to 10 bit - maybe even 12 to cover extra HDR range, i.e. higher absolute brightness when the display supports it. (Or go with a logarithmic encoding, in which case even 8 bit might be enough...)

TwoBit 6y ago

Then they ran the test wrong. It's easy to construct 10 bit color images which are clearly discernable from 8 bit.