attention color whizes: non-typical sRGB/RGB/ProPhoto question

[bjanes]

> It does seem to me that joofa has a point, but it's most likely not relevant for everyday use.

If you look at how proofing is done absolute colorimetric intent is in use there for certain cases. The goal of proofing is to mimic behavior of one media using a different media. Consider the real life scene to be presented to the eye observing it in person as one media, and the presentation of that scene on a display or on a print as the other media. Now what are we going to use as the rendering intent if we want to recreate the original impression of the scene?

Bruce Fraser had a good post on this subect 10 years ago.

Regards,

Bill

[MarkM]

I couldn't find a utility that would let me compare gamuts based on two white points in XYZ space. (I thought maybe the geometry would be a little easier to grasp than LAB plots). I had a little free time today so I fired up the python interpreter and blender to take a closer look at what I think is the issue of this thread.

Attached are two renderings of the parallelepipeds in XYZ space comparing the Adobe98 profile and the ProPhotoRGB profile.

In both the black point is the lower left vertex and white(s) are in the upper right.

The first image with the Adobe space completely contained is converted to XYZ using a matrix based on the Bradford-adapted primaries (e.i. the primaries in the actual AdobeRGB1998.icc file). I have not added any additional chromatic adaptation. The white points are equal when you do this.

The image that shows the Adobe space extending out of the ProPhoto space in the Z direction uses the matrix from Bruce Lindbloom's site labeled AppleRGB   D65. (The primaries that create this matrix correspond to the chromaticities in the Adobe1998 spec, but are not the primaries found in the .icc file). If you had nothing other than the .icc file I'm not sure why you would use this matrix—you would have to look at the media white point  in the .icc file and reverse the Bradford adaptation to get this matrix. Apparently that's what the Apple CMM does though when you set it to absolute colorimetric intent, so someone thinks it's a good idea.

As you can see in the second image, when you use these primaries, the white point is pulled in the Z direction and pulls the blue, magenta, and cyan vertices out of the proPhoto space. The line between the black point and the white points is now a different vector, which means greys in one aren't neutral. Which grey isn't neutral depends on your adaptation state.

It's tempting to think (as this thread has amply demonstrated) that the second image shows a difference in gamut. But another interpretation is that it shows the effects of chromatic adaptation on color perception. If you put yourself in the shoes of two different observers, one adapted to D65 light and another to D50 light, you can make some sense of the second chart. The D65 observer has become less sensitive to blue relative to the D50 observer due to the extra blue in D65 light. The D65 white point looks white and the D50 white point looks yellow to him. The blue primary in the Abode profile from here looks like a nice saturated blue. Now, if we look at if from the D50 observer, things look much different. The D50 white point looks white, of course. The D65 white point looks impossibly white/blue, whiter that white-like what laundry detergents promise. The blue primary of the adobe profile is super-blue, it's much bluer to you than it looks to the D65 viewer. There's a color you can see which looks just like the color the D65 viewer saw when looking at his blue primary. That's the Adobe98 blue primary in the first chart, which is happily within the gamut the Prophoto Space. That color to you looks exactly like the Adobe blue primary in second chart looks to the D65 viewer. One might say they are the same color even though they have different coordinates in XYZ space.

Clearly this idea represents a difference of opinion that is irreconcilable for some.

If anyone is interested, I'm happy to email the blender files that create these renderings. (Blender is free, but a pain in the ass to learn). I'll also send the python scripts that modeled the actual shapes and coordinates if anyone wants to play with them—just send me a note.

Below: ProPhoto RGB (the semi transparent shape) and Adobe1998 (the opaque shape) in two interpretations.

[ejmartin]

Is it just the perspective, or are the sides of the parallelepiped not parallel in the plot?  They should be.

[Eric Myrvaagnes]

MarkM,

Thank you for posting both the illustrations and the explanation.

I have followed this thread from the beginning, hoping to learn something about color management, but most of what I have seen has been unlabeled plots that don't even state clearly what we are looking at, or sets of numbers with no meaningful context. I have been hoping someone would actually explain what we are seeing.

If those who have claimed that parts of AdobeRGB lie outside of Prophoto would try to provide a similar level of explanation of what their plots or numbers are about, I think they might stand a much better chance of gaining more respect, at least from me.

Eric

[Graystar]

Clearly this idea represents a difference of opinion that is irreconcilable for some.

It’s a nicely written appeal, but I doubt it will pique the interest of anyone who would continue to respond to my posts even though I was clearly spoofing Joofa (come on!)

I think there are some very basic ideas that are being missed.  First, colors do not exist.  More than anything, color is a figment of our minds.  When two different spectral distributions have the same color response, then they ARE the same color.  It is the response that defines the color...not the light used to produce it.

Second, CIE color spaces are emissive spaces, and color response is dependent on both the tristimulus values and the light used when executing those values (the Standard Illuminants.)  That’s why Adobe RGB Blue will convert to different combinations of tristimulus values and illuminant...they all create the same blue, they just use different spectral distributions to do it.

It should be clear to anyone who knows this information that if you keep the tristimulus values but change the illuminant, you’ve changed the color response (and hence, the color.)  However, that bit doesn’t seem to matter to some people.  As you say, it’s irreconcilable.

[digitaldog]

What Mark’s graph brings to the party for me is the illustration that the gamut doesn’t get larger, there’s simply a significant shift. This backs up the analysis ColorThink Pro provides of the gamut volume differences between Adobe RGB (1998) and ProPhoto RGB (1,205,502 vs. 2,548,220). Using no adaptation, which many here say is the wrong use of these types of display profiles, pushes blues of Adobe RGB (1998) outside the gamut of ProPhoto. Doesn’t make it larger (as claimed in Post #5):

It does not seem like that Prophoto RGB is always wider than Adobe RGB. It seems that saturated colors around blue Adobe primary might clip in Prophoto RGB in standard specification of these spaces. For more information please see:

http://forums.dpreview.com/forums/read.asp?forum=1018&message=37330104

We have an artificial, (some agree incorrect use of these two profiles), showing a theoretical “graph” without full explanation that attempts to back up this theory from post #5. One could say that if you take update the blue primary in Adobe RGB (1998), you can make it do the same thing (after which, someone could say its not Adobe RGB (1998) anymore), and another 30 pages of gobbledygook could be written.

The proponents of the theory still have not provided a means of or the reason to make the shift Mark illustrates. Bruce makes the consequences of doing so clear in his quote.

[sandymc]

All,

We've known what the "answer" is here since page 2 of this thread - simplistically put, some folks don't agree with the orthodox approach to chromatic adaptation in current color management systems. Or perhaps even the need for it.

Which is fine, and the discussion has given us some useful and memorable posts.

But now it's time to agree to disagree and move on - short a coherent, concrete proposal as to what the alternative approach is, this conversation is going nowhere.

Sandy

[tgray]

One could say that if you take update the blue primary in Adobe RGB (1998), you can make it do the same thing (after which, someone could say its not Adobe RGB (1998) anymore), and another 30 pages of gobbledygook could be written.

I think Iliah actually made that point, or a similar one, earlier in this thread, with comparing Prophoto RGB AbsCol with a chromatically adapted version of itself to a bluer color temp.

Nice write up Mark.  Maybe I'm missing something, but that's more or less what I got out of Joofa's original post.  It's nice to see that you took the effort to explain things a bit more thoroughly than Joofa has been willing to do, but it is essentially the same thing as his example with the four cases.

I just don't see what all the hubbub is about. In his original post, he claims that AdobeRGB has blues outside of Prophoto's gamut when you don't chromatically adapt it to D50.  That was it as far as I can read.  And it seems most of you have actually agreed with it., minus the commentary that 'you'd never not use chromatic adaptation' or other objections.  

Mark, you have a big paragraph at the end of your post talking about an alternative interpretation.  A nice description, but again, doesn't this sum it all up:

AdobeRGB has blues outside of Prophoto's gamut when you don't chromatically adapt it to D50.

All the extra discussion, including Bruce's comment, are encapsulated and implied with the "don't chromatically adapt it" part of the statement.  There's a warning right there that if you might not be preserving the perceptual relationships of the color*, since you aren't applying chromatic adaptation.  We usually want our neutrals to be neutral.  And you won't be getting that in this example.  Etc. It seems like to me it's really all there in that sentence.

*excluding proofing and some of Iliah's examples

[joofa]

It's nice to see that you took the effort to explain things a bit more thoroughly than Joofa has been willing to do, but it is essentially the same thing as his example with the four cases.

In an incorrect representation. But there is no point in beating the dead horse. Please see below.

BTW, I think you are being a little unfair here. I have written detailed posts to illustrate some fundamental misunderstandings and how to do proper evaluation of colorimetric entities.

I just don't see what all the hubbub is about. In his original post, he claims that AdobeRGB has blues outside of Prophoto's gamut when you don't chromatically adapt it to D50.  That was it as far as I can read.  And it seems most of you have actually agreed with it., minus the commentary that 'you'd never not use chromatic adaptation' or other objections.  

I think it is time to move on. Iliah has given a direction: When not to consider chromatic adaption. I think it is better to concentrate here now as this is a topic I also want to learn.

Sincerely,

Joofa

[tgray]

In an incorrect representation. But there is no point in beating the dead horse. Please see below.

Ok.

BTW, I think you are being a little unfair here. I have written detailed posts to illustrate some fundamental misunderstandings and how to do proper evaluation of colorimetric entities.

You did write detailed posts explaining your actions later.  Your original one was a bit terse which lead to about 6 pages about how "AdobeRGB can't have a D50 white point, because it's then not AdobeRGB."  I think you didn't imply that but people got caught in the semantics of the terminology.  It probably could have been avoided with a bit more explanation initially.  However, folks here have CLEARLY had some issues with what you wrote and how you've presented it.  Feel free to call me a 'bit unfair' for recognizing that if you choose.

As far as when not to use chromatic adaptation and moving on, it might be time to start a new thread?

[joofa]

Ok.

You did write detailed posts explaining your actions later.  Your original one was a bit terse which lead to about 6 pages about how "AdobeRGB can't have a D50 white point, because it's then not AdobeRGB."  

What was that on the link below, right there on the 2nd page, which illustrated the confusion many were having including Digital Dog and MarkM:

http://www.luminous-landscape.com/forum/index.php?topic=49940.msg412251#msg412251

Sincerely,

Joofa

[tho_mas]

What Mark’s graph brings to the party for me is the illustration that the gamut doesn’t get larger, there’s simply a significant shift.

and this is maybe one of the (initial) reasons for the long debate.... simply sounds "strange" that AdobeRGB is wider than ProPhoto... given that these are artifical color spaces that you actually always use relcol.
Would have been a different debate if Joofa looked at the old D65 sRGB and the revised D50 sRGB profiles ... and stated sRGB is wider than sRGB.

[joofa]

Would have been a different debate if Joofa looked at the old D65 sRGB and the revised D50 sRGB profiles ... and stated sRGB is wider than sRGB.

As impolite it might sound, we would have a different debate if some of the "color experts" had a better idea that what is a color space. Sorry for being a little politically incorrect. I apologize for the bluntness of language here. But you had given me no choice. Beating the same thing after 33 pages!

Joofa

[Graystar]

What was that on the link below, right there on the 2nd page, which illustrated the confusion many were having including Digital Dog and MarkM:

http://www.luminous-landscape.com/forum/index.php?topic=49940.msg412251#msg412251

From that post...
"Instead of a reflector, suppose there is a source that emits two colors A = [0.1881852  0.0752741  0.9911085] and B= [0.14922403  0.06321976  0.74483862] in XYZ, which are, in absolute terms, two different colors."

This is your fundamental misunderstanding.  Besides the fact that colors aren't emitted, the color response of the tristimulus values depends on the illuminant.  With the same illuminant, the two sets of values invoke different color responses. With different illuminants, the color responses are the same.  There’s nothing here that can be called “absolute.”

Color exists in the mind only...not in the light.  Color has no physical properties.  Color exists only as a response to light.  That’s what you have to realize.

[tho_mas]

I apologize for the bluntness of language here

not a problem. understandable.

[MarkM]

Is it just the perspective, or are the sides of the parallelepiped not parallel in the plot?  They should be.

Good eye Emil. The renderer does apply a perspective view, but not enough to account for that. I had a left over gamma exponent in the wrong spot that was skewing the second image. Fix is attached. It's a significant difference, but the main idea should still be the same.

that AdobeRGB has blues outside of Prophoto's gamut when you don't chromatically adapt it to D50.

I guess I would sum up my view as yes and no. I'm much more in agreement with Graystar that XYZ coordinates without reference are not colors, blues or otherwise. And I think it's more than just a matter of semantics.

The XYZ space provides a very elegant model for seeing adaptation. If you imagine standing on the white point like it's the bow of of a ship and looking toward the stern, you see all the color relationships in front of you: you're at the highest point; straight ahead and down by the propeller is the black point, if you look further down things get yellow until you look almost right under your feet to the yellow point. Magenta/red are on your left, blue/cyan are on your right. Blue is above the black point, but still a little below you. If you look up or in the other quadrants, it's is the land of oxymorons, imaginary colors, blacker than black, etc. What's important is not the coordinate you're looking at, but the direction from where you're standing.

So imagine standing on the D65 white point and looking at that contentious blue point in the second image. You keep staring at the blue point as someone cranks a lever lowering you and the colors around you into the position of the first image until the white point you're on is now D50. The point you're looking at doesn't change color as this happens because you're still looking in the same direction, it still appears to be the same blue but now it fits easily inside the ProPhoto space. At this point you look up at the original coordinates of that blue and it looks totally different. At no point during this exercise did you ever see THAT color blue. That has never been the color we are talking about, it's only a set of coordinates. The color we're talking about, the one you saw in front and below you from the D65 white point, that color is easily reproduced with the prophoto space. It's not out of gamut.

Of course I understand that you can see things the other way. I just think we're talking about two different colors when you do that.

[joofa]

What Mark’s graph brings to the party for me is the illustration that the gamut doesn’t get larger, there’s simply a significant shift.

Hi,

According to my calculation the gamut of Adobe RGB (D65) is larger than that of Adobe RGB (D50). This is a very preliminary calculation and I have to verify it further yet.

Sincerely,

Joofa

[digitaldog]

According to my calculation the gamut of Adobe RGB (D65) is larger than that of Adobe RGB (D50). This is a very preliminary calculation and I have to verify it further yet.

One of your issues is language. There is no such thing as Adobe RGB (1998) D50 as was discussed a very long time ago. There’s some color space with Adobe RGB (1998) primaries and a D50 WP you are comparing (god knows how). The gamut calculations from ColorThink clearly produce a report that shows Adobe RGB (1998) is smaller than ProPhoto RGB. If you think its important, I’ll build a variant (JoofaRGB if you will), with D50 WP and ARGB primaries and see what ColorThink now reports, at least between Adobe RGB (1998) and JoofaRGB.

This language issue goes back to your first post in this thread:

It does not seem like that Prophoto RGB is always wider than Adobe RGB

I think what you could have originally said was “It does not seem like Adobe RGB (1998) is fully contained inside ProPhoto RGB, under a condition which probably isn’t possible outside of a web calculator, here you see blue fall outside ProPhoto RGB”. Would have saved a lot of posting between various parties!

Then like above, you started talking about Adobe RGB (1998) with a D50 white point which again isn’t possible, at least it would not be Adobe RGB (1998). We then got into the ideal of joofaRGB.

Then later still, thanks to others, we learned that its not Adobe RGB (1998) with a D50 WP that produces the results you report but rather viewing the two color spaces without chromatic adaptation of which Mark’s last set of posts shows clearly (where most, certainly I agree) that blues fall outside ProPhoto RGB but no, the gamut isn’t larger. Big difference!

[joofa]

Digital Dog,

We are running in circles. I think we should now move forward to the implications of not using chromatic adaption.

However, I am putting the terminology as used by me here so that there is no confusion.

Adobe RGB (1998) = Adobe RGB (D65).
Adobe RGB (1998) w/ D50 white point = Adobe RGB (D50)
Prophoto RGB = Prophoto RGB (D50).

Hope that helps.

Now lets move on.

Sincerely,

Joofa

[tho_mas]

I've made an AdobeRGB profile in Photoshop adapted to D50 based Br. Lindblooms adapted values.
This is slightly larger than the regular D65 profile.
I just don't know whether or not Photoshop is the appropriate tool ...