Why Do We Need Colour Working Spaces

[Mick Sang]

Hello,

A question which has me stumped currently is why we need subset colour working spaces i.e.While I understand that colour working spaces are subsets of colour designed to serve a particular group of processes, why can the spectrum locus not suffice as the one and only colour space to which all software and equipment refers? I hope someone in this forum will have some thoughts on this.

Thanks,
Mick

[Manfred M]

Simply because different tools (screens, printers, etc) can only reproduce so many distinct colours.

The sRGB colour space is effectively what your typical computer screen of the 1996 vintage was capable of reproducing (and remains the defacto Internet colour standard). Your typical current low end computer screen can handle the sRGB colour space and no more.

Some of the highest end colour screens currently on the market can do pretty well the full AdobeRGB colour space. Your typical three colour (CMY) + black (K) printer can reproduce the CMYK colour space.

[Simon Garrett]

By "spectrum locus" what co-ordinate system do you mean?

As most images are represented in RGB colour spaces of some sort, I guess Adobe and others designing editors etc find it convenient to use RGB for most purposes. A principle requirement for a working space - the co-ordinate system in which one does the processing on image data - is that it should have a gamut as large as any colour you ever want to represent (and preferably larger, to allow for intermediate results that might overflow the range).

ProPhoto RGB is the widest commonly available RGB colour space, which would generally be my choice.

Am I missing something here? What co-ordinate system would you suggest for internal calculations on image data?

[Mick Sang]

[QUOTE=Simon Garrett;516131]By "spectrum locus" what co-ordinate system do you mean?

Thanks for this. I appreciate the reply. By "Spectrum Locus," I am referring to the gamut of human vision.

I understand that sRGB, Adobe Wide Gamut RGB, ProPhoto RGB and all RGB working spaces capture or encompass sections of the Spectrum Locus in order to accommodate some particular set of functions. But, I wonder why there are so many so called working spaces when the Spectrum Locus encompasses all of human vision and hence, all of the working spaces. So, why do we not refer to it directly rather than translating from it to RGB or some other space and back again. Is it purely for some convenience? I have trouble with that answer because ProPhoto captures an extremely large chunk of the Spectrum Locus and actually exceeds it by including colours that we can not see.

Mick

[Simon Garrett]

There needs to be some numerical co-ordinate system in which to represent that gamut of human vision. Most co-ordinate systems don't exactly match human vision (no more and no less), but there are some that include most of it and more, such as ProPhoto RGB. See https://en.wikipedia.org/wiki/ProPhoto_RGB_color_space.

ProPhoto RGB covers virtually all humanly perceptible colours. Those outside are ultra-saturated greens and magentas which are rare in nature. Also, the CIE diagram is not perceptually uniform. Effectively, the area of highly saturated colours on the diagram is stretched. The areas of colour outside the ProPhoto triangle are a fairly insignificant part of our human gamut of perceptible colour.

ProPhoto RGB also includes non-colours, as you mention. Those points within the triangle on that Wikipedia link but outside the CIE horseshoe are not colours. It's not that we can't see them; they don't exist. They don't correspond to any wavelength of light or possible combination of wavelengths, they're mathematical anomalies.

You mention "translating from it [human spectral locus] to RGB or some other space and back again", but I don't think that happens. Most image devices generate RGB data, most stored images (including raw images) are represented in some form of RGB data, most output devices (monitors, printers etc) want RGB data, so it's probably easiest to use RGB for processing. I don't think RGB co-ordinate systems can exactly match the human locus of colours - to include most or all human colour means also including non-colours within the number range, as does ProPhoto RGB, but I don't think that matters.

The main thing is that the working space should include all colours you might ever need, and I think ProPhoto RGB does that.

[Krawuntzel]

[QUOTE=Mick Sang;516141]

So, why do we not refer to it directly rather than translating from it to RGB or some other space and back again. Is it purely for some convenience? Mick

Why do you not ask Adobe?:
Charles Geschke, John Warnock (Board of Directors),
Shantanu Narayen, CEO
and go on to ask the sales managers of Canon, Nikon, Sony and so on: finally arriving at printer manufactures.
Seems to be convenience photography food
Erwin

[Simon Garrett]

[QUOTE=Krawuntzel;516288]

Why do you not ask Adobe?:
Charles Geschke, John Warnock (Board of Directors),
Shantanu Narayen, CEO
Seems to be convenience photography food
Erwin

Not sure what's going on with the quoting system, as the question "...why do we not refer to it directly..." came from Mick Sang, not me.

I wouldn't expect a knowledgeable answer about working spaces from company board members or sales managers.

Eric Chan (designer of Adobe Camera Raw) might be better.

However, if you were designing software to take RGB data as input, process it and then produce RGB data as output, what working space would you choose?

[Krawuntzel]

I wouldn't expect a knowledgeable answer about working spaces from company board members or sales managers.

Maybe not as knowledgable as you would like it, but you might get the real life answer: our customers want that; and: we sell it to their likes. And we do not fuss around with something even better, which is costly to produce and will not sell in the short run (or something along these lines).

However, if you were designing software to take RGB data as input, process it and then produce RGB data as output, what working space would you choose?

If I were in this situation (on my own time), I would use the largest available
Erwin

[Manfred M]

Let's look at the history of a few well known colour spaces. First of all, the baseline was set by the Commission internationale de l'éclairage (ICE), in 1932, (English translation: International Commission on Illumination, which in spite of having a French name is actually based in Vienna, Austria). This colour space represents all colours that are what was determined to be in the human vision range (whatever that means). All current colour spaces are compared to how closely that they come to ICE 1932.

HP and Microsoft introduced sRGB in 1996 in support of what was then the requirement for a standardized way of displayig colours on the technology that was available in computer screens back then. At the time HP was a major player in the PC world and was the #1 producer of personal computers. This colour space was adopted by others and became the defacto standard. It remains the most commonly used colour space, even today as this has become the "lowest common denominator" for images on both the internet and "commodity"; i.e lower end computer screens. If I recall correctly, it covers around 10% of the colours that are visible to humans. Unlike the CIE 1932 standard, the colour space had to be described in mathematical terms so that a computer could quickly calculate and display them.

In 1998 Adobe introduced AdobeRGB which addressed another problem. Adobe software not only handles the work done on a computer screen, but also extends to the publishing industry that produced printed output on offset presses (and to a lesser extent, the colour inkjet printer). the AdobeRGB colour space was developed to unify the work done on sRGB computer screens that was output on CMYK printers. Again, if I recall correctly, that while AdobeRGB was a major step beyond the number of shades / colours that could be reproduced using the sRGB colour space, it was still only around 25% of the total visible colours. The CYMK colour space is quite small, and while there are standards, these are highly dependent on the paper type and inks used and there are even geographical differences (different "standareds" in North America, Europe, Japan, etc.).

This of course means that these two colour spaces are still limited by the fact that there are other colours that humans can see that cannot be reproduced by these two colour spaces. Adobe came up with an extended AdobeRGB colour space and Kodak introduced ProPhoto that is even wider than the other colour spaces. This colour space is supposed to cover all colours and shades that can be seen by humans, but also includes some colours that are "not real". It seems to have become the defacto "standard" if for no other reason than it is supported by Adobe Photoshop.

So, there are many reasons for why there are different colour spaces, and why they are still in use and relevant today. They were developed to meet specific (commercial) needs. Ultimately, the devices that are being used drive the need for colour spaces, so while there are unified colour models, they have practical limitations, but let's not get into this here...

[xpatUSA]

Why do you not ask Adobe?:
Charles Geschke, John Warnock (Board of Directors),
Shantanu Narayen, CEO
and go on to ask the sales managers of Canon, Nikon, Sony and so on: finally arriving at printer manufactures.
Seems to be convenience photography food
Erwin

There might be a language difficulty but this looks to me like unwarranted sarcasm, perhaps needing the attention of a Moderator.

[Simon Garrett]

If I were in this situation (on my own time), I would use the largest available
Erwin

I'd agree, which for me would mean ProPhoto RGB. But would you use an RGB working space? To me that would seem the logical choice, but I wonder if someone might say that computationally there could be advantages in using an LAB space, for example.

[Manfred M]

There might be a language difficulty but this looks to me like unwarranted sarcasm, perhaps needing the attention of a Moderator.

Yes and no Ted, as I suspect that this may not be far from the truth. There is nothing nefarious here, companies do things that are right for them and the fact that we are using them today suggests what they did seems to work.

The common colour spaces were developed for commercial reasons. When I was researching this subject some time ago, the AdobeRGB specs were defined by Thomas Knoll (the guy that "invented" Photoshop). He made some fundamental errors in the developing the specs, but it is still a very widely used colour space.

[xpatUSA]

Some of pedantry, Manfred, if I may?

Let's look at the history of a few well known colour spaces. First of all, the baseline was set by the Commission internationale de l'éclairage (ICE), in 1932, (English translation: International Commission on Illumination, which in spite of having a French name is actually based in Vienna, Austria). This colour space represents all colours that are what was determined to be in the human vision range (whatever that means). All current colour spaces are compared to how closely that they come to ICE 1932.

That date is usually quoted as 1931.

In 1998 Adobe introduced AdobeRGB which addressed another problem. Adobe software not only handles the work done on a computer screen, but also extends to the publishing industry that produced printed output on offset presses (and to a lesser extent, the colour inkjet printer). the AdobeRGB colour space was developed to unify the work done on sRGB computer screens that was output on CMYK printers.

As you mention later, there are other Adobe color spaces so we should write "AdobeRGB (1998)".

Adobe came up with an extended AdobeRGB colour space and Kodak introduced ProPhoto that is even wider than the other colour spaces.

Kodak introduced their "ROMM" output-referred color space which, only later, became known informally as "ProPhoto".

This colour space is supposed to cover all colours and shades that can be seen by humans

No, neither Adobe Wide nor ProPhoto cover all the colors. Some blues, cyans, are excluded.

It seems to have become the defacto "standard" if for no other reason than it is supported by Adobe Photoshop.

Don't get me started . .

[Manfred M]

Some of pedantry, Manfred, if I may?

To paraphrase a former boss (also an engineer, by the way).

Arguing with an engineer is somewhat akin to mud-wrestling with a pig. After you've been at it for awhile; getting hot, sweaty and frustrated, you discover that the pig is actually quite enjoying itself...

Part of the problem with defining the range of human vision is establishing what that actually means. In the work I've done on designing for specific populations of users (35+ years of this type of work), the approach our design teams have always followed was to ensure that anyone from the 5th percentile to the 95th percentile of the user group would define the ergonomic design parameters of the project. I suspect that this was not followed by the CIE when they set up their standard.

[xpatUSA]

Part of the problem with defining the range of human vision is establishing what that actually means. In the work I've done on designing for specific populations of users (35+ years of this type of work), the approach our design teams have always followed was to ensure that anyone from the 5th percentile to the 95th percentile of the user group would define the ergonomic design parameters of the project. I suspect that this was not followed by the CIE when they set up their standard.

I'm not so hot on statistics, Manfred. Could you tell us what a 'percentile' is, please? (I do have a vague memory of two standard deviations each side of the mean as encompassing 90% of a normally-distributed population).

CIE have released many standards:

In the original CIE stuff they used a standardized average-of-their-subjects called the "2-degree observer". As I suspect you know, later (1964) there came along a "10-degree observer". And then they started work on means of presenting information in a friendlier manner that the 1931 xyY chromaticity diagram - that's the one that fools people into thinking that the AbobeRGB (1998) color space gamut is much bigger than that of sRGB! For example, both the little-known CIELUV and the better-known but less-understood CIELAB present the various color space gamuts to the eye in a manner more in keeping with our eyesight, especially in terms of differences between colors.

Not going to mention CIECAM today . . .

[davidedric]

And the women in the uplands digging praties,
Speak a language that the strangers do not know.

Well, I'm lost anyway, and I'm certainly done mud wrestling with engineers

[Manfred M]

I'm not so hot on statistics, Manfred. Could you tell us what a 'percentile' is, please?

d

When I look at both the 5th percentile and 95th percentile, these terms merely mean that 5% of the total population is outside the control limit at either end. If I look at the North American population, the 5th percentile female is 5' 0", which means that 5% of women are shorter than 5' 0". The 95th percentile man is 6' 2" which means there are 5% of the North American population that are taller than 6' 2". You can do this for all kinds of other anthropometric measurements; hand width, arm length, visual acuity, etc.

[loosecanon]

I find I get problems if I work and perfect (as far as possible for me) in prophoto and then have to convert to sRGB for use as a digital file by someone else. I have found significant colour shifts that are not quick to resolve, so now I have moved back to adobeRGB for editing, and when necessary conversion to sRGB.

[Simon Garrett]

I find I get problems if I work and perfect (as far as possible for me) in prophoto and then have to convert to sRGB for use as a digital file by someone else. I have found significant colour shifts that are not quick to resolve, so now I have moved back to adobeRGB for editing, and when necessary conversion to sRGB.

That's interesting. In theory, I should have thought that with proper rendering from ProPhoto or Adobe RGB to sRGB, then there shouldn't be any difference. I wonder if the difference might be either less-than-perfect rendering (in Photoshop or whatever you're using) or less than perfect profiles?

[Richard Lundberg]

All space conversions in Photoshop are Relative Colorimetric for one thing. You only get Perceptual, Saturation, or Absolute if the printer does it.

As for the spaces everything was 8 bit then, and it's just barely adequate for Adobe RGB.