So, Steve, anything that is outside of that "1 Lab" unique cube is considered a "different color", in ColorThink? Is that your criteria? / Roger -----Original Message----- From: colorsync-users <colorsync-users-bounces+graxx=videotron.ca@lists.apple.com> On Behalf Of Steve Upton via colorsync-users Sent: Tuesday, January 7, 2020 3:59 PM To: colorsync-users@lists.apple.com Subject: Re: 1 billion colors
On Jan 6, 2020, at 1:36 PM, Roger Breton via colorsync-users <colorsync-users@lists.apple.com> wrote:
Well, Andrew, just to push the discussion further, I would *love* to hear what Steve Upton would have to say about “unique colors” since he’s the one that wrote the application: what numerical criteria does Steve use in ColorThink to distinguish among “unique colors”?
Gah! I've been drawn into the angels-on-the-head-of-a-pin discussion again! I agree with all of you... does that help? I have to admit the "There's a color missing, prove me wrong" is pretty damn funny. Good to get a chuckle from this stuff sometimes. My 2 bits? There's an interesting point here about the difference between addressable colors and addressable unique colors. For me (not necessarily using the correct terms) an addressable color in this discussion would be any RGB value that results in a Yxy value that falls within the chromaticity diagram and represents an actual, visible color. This includes any precision level so imaginary 128bit-per-channel RGB spaces would be included as well. There'd be a lot of wasted bits addressing imperceptible color differences but they'd be real colors. This brings me to: Addressable *unique* colors is the idea that perceptual spaces like Lab have a kind of maximum resolution, where we could use a set of coordinates that is no more precise than our ability to see the differences between colors. Each unique color is effectively assigned a coordinate and we're good to go. "But wait", you might say, "color perception is analogue and infinitely variable" (nod to Mr Gnaegy, good to hear from you sir). "A hard-edged coordinate system could never capture it!" I agree that we're analogue beasties but it's not a problem. In our discussion we're talking about trying to address unique colors. If we use such a hard-edged coordinate system there's nothing saying it can't be slid around / re-aligned with infinite precision, only that the differences between the coordinates would need to remain the minimum perceptible difference between colors - therefore it still addresses unique colors... If I understand and slightly warp the intentions of our color forefathers properly, CIELAB was intended to be just such a coordinate system, where 1 unit of movement in any direction would be the human perceptual limit of color change. If you added up the "volume" of such 1-unit cubes in a defined area, such as a printer gamut, then you'd have a number representing the number of unique colors available. That's exactly what I do in ColorThink. Unfortunately CIELAB is flawed, or perhaps it's better to say that human color perception doesn't adhere to such a simple model. Newer models have been created such as CIECAM that help address some of these limitations. Suffice to say that there's broad agreement that CIELAB and delta-E 76 are dangerous to use naively, yet remain quite useful and delta-E 2000 helps bring color differencing numbers more in line with human perception. Though I recognized that the concept of a total gamut volume is disputable and somewhat flawed from the start, I thought that it probably still had value and chose to include it in ColorThink Pro. At the very least I thought it would start some interesting discussions! Overall, I think it has been useful, especially in a comparative sense. Are there better ways of doing it? Undoubtedly. Is there agreement on which ways are better? Not much that I can find yet, but work on the science continues and we continue to learn as we go. Those who are attending the Color20 Conference in San Diego next week will be treated to a sneak peak of what's coming in ColorThink 4 later this year - like new color spaces for graphing, advanced gamut comparison functions, and so forth. regards, Steve _______________________________________________ Do not post admin requests to the list. They will be ignored. colorsync-users mailing list (colorsync-users@lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/colorsync-users/graxx%40videotron.ca This email sent to graxx@videotron.ca
On Jan 7, 2020, at 1:36 PM, Roger Breton via colorsync-users <colorsync-users@lists.apple.com> wrote:
So, Steve, anything that is outside of that "1 Lab" unique cube is considered a "different color", in ColorThink? Is that your criteria?
effectively, yes. Though if you "do the math", it becomes clear that 1 Lab cubes don't pack properly so that the centers of each cube are 1 dE apart from all the other centers, *and* that you can't really do that because sphere's don't pack efficiently, etc, etc, etc Steve
On Tue, Jan 7, 2020 at 14:21 Steve Upton via colorsync-users < colorsync-users@lists.apple.com> wrote:
On Jan 7, 2020, at 1:36 PM, Roger Breton via colorsync-users < colorsync-users@lists.apple.com> wrote:
So, Steve, anything that is outside of that "1 Lab" unique cube is considered a "different color", in ColorThink? Is that your criteria?
effectively, yes.
Though if you "do the math", it becomes clear that 1 Lab cubes don't pack properly so that the centers of each cube are 1 dE apart from all the other centers, *and* that you can't really do that because sphere's don't pack efficiently, etc, etc, etc
Steve
Roger, it seems you are thinking about something like packing of MacAdam ellipses... https://en.m.wikipedia.org/wiki/MacAdam_ellipse The MacAdams ellipse describes relative capacity to distinguish color across regions of the locus. Normal individual observers — normal meaning those who you would not define as color-blind — will have varying abilities to discriminate depending on the color. (Is it fair to use the term color in this way? If color is a pure qualia can you have color blindness? How can one be blind to that which for him does not exist? ;) The idea is you pick a coordinate and ask how well an individual, as compared to the population, is able to discriminate difference surrounding this coordinate? And how sensitive is the population to differences across the gamut (word police: "we have you surrounded"). The MacAdam idea leads to a heap or camel or Zeno's paradox: Pick a point and draw a MacAdam ellipse. Within the ellipse, color is considered indistinguishable. Between the neighboring ellipses, color is distinguishable. Pick a point half-way between neighboring ellipses and draw another MacAdam ellipse. Where did the color go?! This may leads me to consider I want an encoding system with a resolution much finer than one standard color difference. Say if I had a system of millions of "colors" nee qualia (oh, wait, strike that, reverse it... Er no wait... waaaahhhg!) and I want a stimulus system that driven by a numerical non-uniformly perceptually distributed integer RGB values to avoid visible quantization artifacts, I may find I want the data format to encode a couple orders of magnitude more stimulus than qualia to ensure I cover the corner cases. Therefore "billions of colors." QED Seeing this MacAdam idea really affected my visualization of the distortions inherent to the XY plot. And cemented my view that Adobe RGB was singled-minded in desire to cover press, with newer DCI being a much more balanced extension of sRGB. As somewhat related aside, DisplayCal has gamut projections in a mode called DIN99, which looks to me like it might be intended to be a very perceptually weighted plot. There's not much about it returned by web search... I admit I haven't asked at the obvious place, over at the DCal forum. I'll post the Dell gamut vs standard gamuts as DIN99 so others can see.
participants (3)
-
graxx＠videotron.ca
-
Steve Upton
-
Wire ~