Too Much Color

45 points by polywolf 8 hours ago on lobsters | 17 comments

majaha | 6 hours ago

fedemp | 4 hours ago

I got 0.039 using my Thinkpad, which does not have the best screen. Noticed that angle of vision surely makes a difference.

technetium | 3 hours ago

Same. I started off on my aging LCD monitor and immediately noticed cases where I could just tilt my head a little and spot an obvious difference. Once I moved the window to my OLED laptop screen, it seemed to matter a lot less, though.

Obligatory score link: https://www.keithcirkel.co.uk/whats-my-jnd/?r=AhogKP_-f847

klingtnet | 6 hours ago

I scored 0.0041, because it was impossible to tell difference at 0.0032 for me. I wonder if my eyesight or the display is to blame?

pointlessone | 5 hours ago

On the fediverse people noted that MacBook screens helped.

From my experience I can say that some hues are harder for me to tell apart. Maybe I have some slight color “blindness” or rather different sensitivity in different parts of the spectrum. So it might be not only absolute distance between colors but also where on the spectrum they are.

Relax | 5 hours ago

We humans absolutely do have varying sensitivity across the spectrum -- we're generally much more sensitive to green, for instance, which is why some screens have more green subpixels than red or blue.

pointlessone | 5 hours ago

Funnily enough I had a lot of trouble in the "olive" territory.

Relax | 4 hours ago

Olives are ok; I think they can be a little overpowering sometimes. Always like a good olive tapenade, though.

Garbi | 3 hours ago

0.0093 I'm gonna go ahead and trust him.

majaha | 4 hours ago

To be honest, my screen isn't amazing and I do wonder how much inaccurate colours or viewing angles matter, as my screens colours can change quite a lot as I move my head around. My strategy was physically scanning my ahead across the screen, trying to use every patch of eyeball and monitor angle.

sebastien | 2 hours ago

You hacked the code, or you have an Eizo screen!

mplant | 3 hours ago

Mine: https://www.keithcirkel.co.uk/whats-my-jnd/?r=AbcdKP__PPDL I think maybe the monitor has weird color calibration

0.0047 on a MacBook Air M3 13" screen. https://www.keithcirkel.co.uk/whats-my-jnd/?r=AdUiKP__v4f3

I would like to know if this is using Display P3 gamut.

bazzargh | an hour ago

Also 0.0047, on a MBP with Dell U2722DE monitor https://www.keithcirkel.co.uk/whats-my-jnd/?r=AdYfJv__t8-M

chinmay | an hour ago

Comment removed by author

pointlessone | 6 hours ago

This sounds overly complicated but maybe I'm missing something. I'd have reasoned that 24-bit RGB has three channels spanning [0,255] so any colour representation with three channels spanning more than 2.4 figures per channel contains excess precision.

In other words, uniquely identifying a 24 bit colour requires about 24 bits of information, regardless of our choice of coordinates. Authors "perceptual limit" is a 20.5 bit representation, which seems reasonable -- I can't perfectly discriminate colours in 24 it RGB anyway.

What mistake am l making?

[OP] polywolf | 36 minutes ago

Two that I can think of:

  1. This author's end goal is CSS minification, so 3dp is the best you can do to achieve "2.4 figures per channel".
  2. OKLAB/OKLCH are perceptually uniform, which almost certainly means they are nonlinear in sRGB. This means we need a bit of excess precision in the former to actually map to all the values in the latter. Maybe not a whole 3dp (haven't done the math so idk, author probably has tho) but again see point (1).