The Mystery Of Tetrachromacy (color vision stuff)

[digitaldog]

I dunno,  it seems like the answer could be obtained pretty easily with a short survey involving an apple, a print, and a group of volunteers.

That much I agree.

Since this is precisely what the average rube thinks his new color munki spyder thing is supposed to do for him, and adding to that the fact that experts become surprisingly evasive when asked these sorts of simple questions, perhaps one can begin to understand my position.

With respect, based on your comments thus far tonight, you seem to be well positioned to speak for the average rube about color management . But then you yourself wrote quite clearly about your crusade, your opinion of color management (with no basis in fact) and your confusion between colorimetry and color perception. My suggestion is, spend your valuable time conducting such a survey. Or maybe get a book on color management (http://www.amazon.com/Real-World-Color-Management-Edition/dp/0321267222) would be my recommendation for you.

[GrahamBy]

The qualities of your display and it's calibration play a role on-line.

So it would seem. I did the "line up the blocks" thing one evening at work and got a highish score, but certainly not 100%.
Then today I looked at the hutchcolor Lab-ramp grey scales on the same un-calibrated monitor: there is some vague pinkness going on over on the right, then one specific step at around 80% density lights up as dark red.

So I think I can now excuse myself for some colour perception failings with this monitor :-)
(The same image on my home screen is resolutely grey...)

[amolitor]

Yes. Let us assume I am ticking all the boxes and doing the work. This is a straightforward question with a straightforward answer, which neither I nor digitaldog know. I would like to know but since I don't use a workflow of the sort described, I am not well positioned to find out.

I submit that it's somewhat surprising that digitaldog doesn't know the answer, however.

[digitaldog]

Yes. Let us assume I am ticking all the boxes and doing the work. This is a straightforward question with a straightforward answer, which neither I nor digitaldog know. I would like to know but since I don't use a workflow of the sort described, I am not well positioned to find out.
I submit that it's somewhat surprising that digitaldog doesn't know the answer, however.

I told you it's not possible to answer rhetorical questions without a lick of color science. Something you appear to ignore or not understand. I can make up an answer while blowing smoke up your rear end but I refuse to do that. It's pointless. Much like your question and your crusades. It still appears you're no farther along distinguishing between Colorimetry and color appearance than you were last week.


Colorimetry and the dE testing is about color perception. It is not about color appearance. The reason why viewing a print is more valid than measuring it is because measurement is about comparing solid colors. Color appearance is about evaluating images and color in context which measurement devices can't provide.

[amolitor]

It's not rhetorical at all.

You need an apple, a color managed workflow to print, and ten friends. I have the first and the last, but not the middle one.

[digitaldog]

You need an apple, a color managed workflow to print, and ten friends. I have the first and the last, but not the middle one.

I can imagine you have the first, wonder about the 2nd and possibility the third. 
Well so much for your testing, unfounded conclusions and crusade.

[amolitor]

You really don't read so good. I said specifically that I haven't got the second. If you won't read all the words of a fairly short post there's just not much point.

[digitaldog]

You really don't read so good.

You don't write very well. 

I said specifically that I haven't got the second. If you won't read all the words of a fairly short post there's just not much point.

A color managed workflow, got it. Yet you admitted originally: Also, more grist for my little personal crusade to the effect that color management is mostly BS.
You're correct about two items thus far in this discussion you've admitted to:
I dunno. That much is rather clear.
2nd is you don't have a color managed workflow to print. Thus the BS crusade based on a process you're unable to accomplish. As I said, stick to writing for beginning photographers and buy yourself a book on color management and attempt to understand the process before going on silly crusades motivated by your own BS. 
Yes, there's just not much point in what you're attempting to tell us.

[amolitor]

I've been having a conversation with a real color scientist, and it's been interesting.

There is population variation out there. The metrics we use in perceptual color spaces agree pretty well with the real world, and are technically achievable, with effort.

So two prints that the metrics say will look the same, will look the same across 'most people' and if the metrics suggest 'small differences' most people will agree. Which is as we'd expect, of course, but what we expect and what is true are not always the same. And, importantly, this level of control is technically feasible.

As for making a print that consistently looks like the apple? No. You can do it under stringent conditions which includes  controlling how you ask the questions as well as the obvious stuff like illumination.

But in practical real world terms the picture of the apple won't match the apple. Which is a bit of an elephant in the room, if you ask me.

Brands go to great efforts to make their single colors look the same worldwide in various conditions, it's very hard, very expensive, and its a single color.

[Jim Kasson]

It's not rhetorical at all.

You need an apple, a color managed workflow to print, and ten friends. I have the first and the last, but not the middle one.

.

You also need a camera that meets the Luther condition (Try finding one of those at B&H), and, the hardest thing is you need to make the viewing conditions identical for the real thing and the reproduction, unless your color managed workflow can compensate for changes in viewing conditions, and I don't know of any that even attempt that across a broad range.

There have been attempts to conduct direct simultaneous viewing of monitor and print that have resulted in very close, but not perfect, matches. The viewing condition matching was very difficult. I am not aware of any similar object to print tests, which would be even more difficult to set up (I think you'd have to limit the observers to one eye at a time to avoid stereoscopic differences for example).

As to the number of subjects, first you'd have to define color normal quantitatively, then devise a test, then test your potential subjects and weed out those that don't meet your definition. 10 would probably be insufficient. There are complaints that the 1931 CIE sample was too small.

When I was doing research of color management, one of the attractive things about it was that the problem as a research project wouldn't be solved out from under me. However, just because color management isn't perfect doesn't mean that it's not useful, any more than General Relativity made Newton's law of gravitational attraction useless.

Jim

[amolitor]

Yep. It's useful, there's a practically infinite amount of chaos to be tamed here, and taming some of it beats taming none!

Thank you for your detailed response!

What seems to be the case is that it's surprisingly hard to articulate even the problems that color management solves or mitigates. The vast majority of people who use color management in some form or another therefore have no clear idea of what problems it's supposed to solve.

It makes things better, in some useful but hard to define way.

Conventional net wisdom is that it's a panacea for all problems. Essentially there is an infinite list of tweaks and dishes you can apply, which provides endless fodder for people on the internet to argue and chide one another.

But you still can't make the picture of the apple look like the apple.

[Jim Kasson]

Yep. It's useful, there's a practically infinite amount of chaos to be tamed here, and taming some of it beats taming none!

Thank you for your detailed response!

What seems to be the case is that it's surprisingly hard to articulate even the problems that color management solves or mitigates. The vast majority of people who use color management in some form or another therefore have no clear idea of what problems it's supposed to solve.

It makes things better, in some useful but hard to define way.

Conventional net wisdom is that it's a panacea for all problems. Essentially there is an infinite list of tweaks and dishes you can apply, which provides endless fodder for people on the internet to argue and chide one another.

But you still can't make the picture of the apple look like the apple.

To see how far we've come (There's a long way to go, and I'm disappointed that many of the things that I thought would get fixed when I left the field twenty years ago haven't) let me take you back to the bad old days before color management (which we called in the early 90s "device-independent color").

Scanner operators would scan 'chromes and (rarely) negs for reproduction on specific presses or at least presses that had been tuned to specific standards. There were more than a 100 knobs on some scanners, which were used to control the CMYK files produced by the scanners. Some of the scanners had only monochromatic monitors, and all the color matching was done by looking at the numbers. The scanner operators were highly skilled. It took years for them to become expert. If you wanted to print on a different type of press, you had to go back and rescan.

Jim

[Jim Kasson]

It makes things better, in some useful but hard to define way.

Here is a look at various ways to frame the issue:

http://blog.kasson.com/?p=12456

The above post includes my paraphrase of Hunt's take on a possible set of color reproduction objectives:



Spectral color reproduction, in which the reproduction, on a pixel-by-pixel basis, contains the same spectral power distributions or reflectance spectra as the original.

    Exact color reproduction, in which the reproduction has the same chromaticities and luminances as those of the original.

    Colorimetric color reproduction, in which the reproduced image has the same chromaticities as the original, and luminances proportional to those of the original.

    Equivalent color reproduction, in which the image values are corrected so that the image appears the same as the original, even though the reproduction is viewed in different conditions than was the original.

    Corresponding color reproduction, in which the constraints of equivalent color reproduction are relaxed to allow differing absolute illumination levels between the original and the reproduction; the criterion becomes that the reproduction looks the same as the original would have had it been illuminated at the absolute level at which the reproduction is viewed.

    Preferred color reproduction, in which reproduced colors differ from the original colors in order to give a more pleasing result.


Jim

[digitaldog]

What seems to be the case is that it's surprisingly hard to articulate even the problems that color management solves or mitigates. The vast majority of people who use color management in some form or another therefore have no clear idea of what problems it's supposed to solve.

You are hardly qualified to speak about color management from your own experiences which you claim you cannot achieve let alone the majority of people who do use color management. You've made a claim of a crusade based on your own admitted ignorance on the topic. You still don't understand nor have you tried to understand the vast differences between Colorimetry and color appearance. You need to spend a lot more time educating yourself on this subject with your buddy the color scientist!

Conventional net wisdom is that it's a panacea for all problems.

Your writings here appear to prove you're the last person who should be talking about conventional wisdom!

But you still can't make the picture of the apple look like the apple.

You can make a picture of an apple appear like an apple.

The metrics we use in perceptual color spaces agree pretty well with the real world, and are technically achievable, with effort.

The sentence as written makes no sense and further proves you don't understand the topic. What metric are you referring to? Perceptual color spaces, such as? Agree with the real world? The effort you yourself claim you are unable to achieve: You need an apple, a color managed workflow to print, and ten friends. I have the first and the last, but not the middle one.

[Jim Kasson]

But in practical real world terms the picture of the apple won't match the apple.

And, in my experience, that's not what most people want. Pleasing color sells better than accurate color.

Jim

[digitaldog]

And, in my experience, that's not what most people want. Pleasing color sells better than accurate color.

Exactly. The scene referred, accurate Colorimetrically 'correct' apple would look awful.The print is output referred. But then the other Andrew is struggling with many of these notions.

[Alan Klein]

And, in my experience, that's not what most people want. Pleasing color sells better than accurate color.

Jim

Since the viewer doesn't see the original apple, as long as the color in the photo appears in a range that looks like an apple, the viewer will accept the color in the inage.    There's nothing to compare too.  That's how I adjust colors from positive chrome film in post processing after scanning them. I never try to match the colors from the film.  I adjust for what looks best to my eyes assuming that other people's eyes are more or less like mine.  Saturation and contrast I adjust to my taste keying off the personality of the original film. (I use mainly Velvia 50 with its high saturation).  Of course then capturing what I see in a monitor onto a print is another matter that is best left to another topic. 

Note: (Flesh tones require closer accuracy because the human brain is very discriminatory with these colors).

[amolitor]

The idea of 'accurate' color in the sense of the apple color matching the apple is certainly quite modern and wrong headed. Most people who buy color management gear think this is what they are buying, however. Look around the internet some, and this is what you will find.

Mainly because people like digitaldog decline to translate their technical language into anything lay people can understand, so people project their own problems onto the solution and hope for the best. Color perception being wonderfully plastic, I dare say they often see their problems being solved. Indeed, I have observed exactly this.

Making it look nice what people ought to be shooting for, yes. The question then becomes 'how much color management stuff do I need to make that happen?'

It all comes back around to 'what problem, exactly, am I solving with this color munki I just bought?' And the answer is remarkably murky. If you're charging real money you should at least be able to explain what you're selling.

[Jim Kasson]

Note: (Flesh tones require closer accuracy because the human brain is very discriminatory with these colors).

In my testing, C1 and Lr default flesh tones are systematically in error in the red direction.

http://blog.kasson.com/?p=12573

In the distant past, I found that was the case with many 'chrome films.

My experience is that people are less tolerant of chromaticity errors along the gray axis than anywhere else.

Jim

[Alan Klein]

Jim:  I'm sure you're right.  There are certainly issues with flesh tones in films like Velvia and Ektar so you should shoot Portra.  The point I was making in my previous post was to forget trying to match the original object color.  Trust the perception of your eyes and brain.  Unless you have some abnormality, most people will agree with your interpretation of colors.  There's no point trying to match the original color unless you're doing pictures of clothes or other products where the the manufacturer needs the colors to match what they're showing let's say in a brochure.   Of course the "munki" is important to make sure your monitor is initially calibrated to "real" base-line color.  Otherwise your eyes will adjust to color parameters that won't match what is "seen" in other monitors.