Color management myths and misinformation video

[MarkM]

Mark, I understand your point and I agree that those different triplets in sRGB map to a single triplet in Adobe RGB. In my monitor, which is calibrated but not a sophisticated one, just a regular one, I cannot see any difference between them in sRGB. I look for the DeltaE between the two colors and the result that I get is 0.0211 (This might be not correct, I'll appreciate if anybody could confirm) which by definition implies that both colors cannot be differentiated

If the issue was precision, we could just move to tiff 16 bits or if that was not enough, then some high end application would handle 32 or more bits per channel, but no real need for it has been proved.

That's the ∆E value I got as well. And yes, I would not call these two values distinguishable colors regardless of the monitor. But I would call them distinct color values and often having more precision that we can see is helpful, especially if you plan on making any adjustments to the image. Again, I think it's a semantic argument and that we essentially agree.

I also think after reading through this thread that Andrew is going to have his work cut out for him.

[smthopr]

fellow color geeks...

I think the confusion in this thread illustrates why this whole subject can be impenetrable to mere mortals.

And, just to add to it, I really don't like the idea of saying that aRGB has more colors than sRGB. Because in fact, it has the exact same number of RGB combinations. And each RGB combination is a unique color value, mathematically speaking.  When you start saying that some values are so close, as to be the same color, it really confuses the whole concept by bringing perception into muddy the idea.

I would suggest maybe you start with something like this:  Assume two color spaces, space "A" and space "B" where each color space consists of 4 colors.  Color space "A" has it's most saturated colors a little bit less saturated than the most saturated colors in color space "B"...

But both spaces contain ONLY 4 colors.  This is something everyone can see without complex 3d graphs.  Extrapolate from there to the spaces with millions of colors.

To me, sorry to say, when you say that Adobe RGB has more colors than sRGB, it's sounds a bit like that fellow who started this whole mess in the first place  

[Slobodan Blagojevic]

... To me, sorry to say, when you say that Adobe RGB has more colors than sRGB, it's sounds a bit like that fellow who started this whole mess in the first place 

And yet none of us could, so far, come up with an alternative, better, succinct way of saying it. The best we've come so far falls somewhere between "can have more" and "more space between pixels."

[Slobodan Blagojevic]

...it has the exact same number of RGB combinations....

Ok, that sounds promising. If so, what makes them different?

[fdisilvestro]

1.  Is it possible that within one of the images we now observe least two patches that can no longer be distinguished?  If so, is this more likely to occur in the sRGB image or the aRGB image?

2.  When comparing individual patches between the two images, will we see differences in some (but not all) patches?

3.  If, between the two images, some patches are distinguishably different,  is this confined to the patches which were outside the sRGB gamut?

Hi, I'll try to answer the questions, others might have different views.

First there some assumptions to be made before being able to give an answer:

- The camera color array satisfies the Luther condition (that's an almost impossible to start with anyway)
- The illuminant is a perfect D65 (another almost impossible)
- The wide gamut monitor ecompassess the full Adobe RGB (another unlikely one)
- Whoever is going to observe the results has obtained a 100% score in the x-rite online color test challenge

So in that perfect world:
1) Yes, it is a possibility, but will depend on how those out of gamut colors map into sRGB

2) If all original patches are different and inside Adobe RGB you should see them also different in the Adobe RGb output. You might or might not see similar patches in sRGB. What you will see in the case of the out of  gamut patches is that their color is different than the original target.

3) Most likely yes, patches that differ between one and the other results should be those out of sRGB

Now, due to the fact that it is almost impossible to satisfy the assumptions, results may vary

[Slobodan Blagojevic]

...It's a container...

If so, what happens when you fill that container with water? Then I guess Adobe RGB "container" contains more water than sRGB one?

[fdisilvestro]

I really don't like the idea of saying that aRGB has more colors than sRGB. Because in fact, it has the exact same number of RGB combinations. And each RGB combination is a unique color value, mathematically speaking.  

Please, FORGET about the numbers, do you want more? What if I develop a software that handles 1 billion bits per channel? Will you have more colors because you have more mathematically different color values?

The numbers are just tools, This is signal processing, just use as many digits as you need. There can be more identifiable colors in Adobe RGB than in sRGB.

[GWGill]

I think the confusion in this thread illustrates why this whole subject can be impenetrable to mere mortals.

And, just to add to it, I really don't like the idea of saying that aRGB has more colors than sRGB. Because in fact, it has the exact same number of RGB combinations.

The problem is right at the beginning. Don't talk about "number of colors" if that is not the concept you want to convey. Say "wider gammut", the correct technical term, and illustrate what that means in being able to reach more vivid colors with AdobeRGB.

[Slobodan Blagojevic]

The problem is right at the beginning. Don't talk about "number of colors" if that is not the concept you want to convey. Say "wider gammut", the correct technical term, and illustrate what that means in being able to reach more vivid colors with AdobeRGB.

Saying "wider gamut" just shifts the burden to defining then "gamut." Besides, "more vivid colors" simply means more colors.

[MarkM]

Saying "wider gamut" just shifts the burden to defining then "gamut." Besides, "more vivid colors" simply means more colors.

If I give you a 4x4 inch square, how many circles can it contain without overlapping? Can it contain more if it's a 5x5 inch square?

It depends what I mean right?

If by circle I mean something that you can visually distinguish from a dot, then you might be able to figure out a finite answer by figuring out the threshold where circles start looking like dots. In this case a 5x5 inch square will hold more of those circles all other things being equal.

But if I mean 'in theory' how many circles can the square hold, it's infinite right? I can always inscribe a circle inside the smallest circle to add one more. Now which square holds more circles?

That's the whole problem here. The math of colorimetry doesn't place a limit on the size of the color point within the space — it has no dimension so an infinite number can fit in the space. It doesn't make sense to talk about how many colorametric points are in a color space.  You can give the point dimension if you define its size with something like a threshold for perceiving differences, but that doesn't seem to be a very useful number to calculate or work with. Bit depth on the other hand is indispensable concept and unambiguously defines how many values we have.  

[fdisilvestro]

While in theory there is no limit on the size of points inside a color space, or in other words that you can have infinite color points, in practice the limit is given by real world limitations such as the capacity of the human perception, noise and the characteristics of the output devices. In practical terms there is no point of using more digits, bits, numbers, than what can actually describe the minimum perceivable difference.

Suppose that you have a scale to measure weight. You can say the the more digits, the more precision, but what if you go to the ridiculous limit of being able to weight every single atom. Would a scale with double the precision be better? In theory maybe, in practice no, because you cannot weight half-atoms.

If you think that 8 bits per channel is not enough, then use 16 bits. Do I use 16 bits? for edit yes, when I'm done and ready for output, 8 bits is all that is required

I have read many times in the Medium format forum how some people have the illusion that because their medium format digital back is 16 bits per channel they will have better color than 14 bits DSLR. Let me tell you, they might have better color but I can guarantee that it is not due to those 2 extra bits. Just look at the noise levels and you can just assign random values to those extra two bits and it would be impossible to tell the difference.

[sandymc]

If I can make a suggestion - the best success that I've had in talking to people for whom "gamut" wouldn't be appropriate is to talk about a "range of colors". So e.g., Adobe RGB can represent a greater range of colors than sRGB.

Talking about "number of colors" generally results in total confusion, because it confuses number of bits in a particular file format with color space.

So actually, in an 8-bit JPEG file, then sRGB and Adobe RGB can represent exactly the same number of colors (depending on how you count colors, which another topic on its own, say 16.7 million odd). However:

• The sRGB version of the file can show a smaller range of colors, but with greater granularity (better color resolution)
• The Adobe RGB version of the file can show a greater range of colors, but with lower granularity

In practice the ability of the human eye to perceive subtle differences in color isn't too great, so the likelihood is that you won't actually see any difference from the granularity.

Sandy

[JRSmit]

Ok, that sounds promising. If so, what makes them different?

The primary difference between sRGB and aRGB in my view is the level of color saturation that can be encoded. The actual encoding in any given color space is limited to the no of bits per primary (in this case R, G, B) used. So it comes down to the color , actually the color objective, you have in an image and if it can be encoded in a color space without falling outside of the color space boundaries, ie its gamut.
This is the technical aspect, and has nothing to do with what our visual capabilites are.

[smthopr]

Please, FORGET about the numbers, do you want more? What if I develop a software that handles 1 billion bits per channel? Will you have more colors because you have more mathematically different color values?

The numbers are just tools, This is signal processing, just use as many digits as you need. There can be more identifiable colors in Adobe RGB than in sRGB.

I understand your argument, but you confuse the concept.  And to be fair, in 8 bit workflows, gamut density does have a real effect.  Our novices are working in 8 bit jpg.  It doesn't matter what software you can develop.  Why make this more abstract?

[fdisilvestro]

I understand your argument, but you confuse the concept.  And to be fair, in 8 bit workflows, gamut density does have a real effect.  Our novices are working in 8 bit jpg.  It doesn't matter what software you can develop.  Why make this more abstract?

Well, I think that even with standard jpg workflows, 8 bit per channel is more than enough even for Adobe RGB, since you can encode more combinations than perceptually different colors, so I cannot see the real effect on gamut density.

Using Sandy's term granularity, sRGB has higher granularity than Adobe RGB, true, but even Adobe RGB granularity is more than what can actually be perceived, so nobody will notice.

I will rephrase my previous challenge: show me two colors that can be perceived by the human visual system as different which require the higher density of 8 bits in sRGB because they will encode to the same value in Adobe RGB.

I bet you that there are no such colors, so the increased density in sRGB is meaningless for practical purposes. My abstraction of using a ridiculous amount of digits was intended to show that numbers are just numbers, not necessarily useful, sorry if it did not help.

[kudzu1804295673]

fellow color geeks...

I think the confusion in this thread illustrates why this whole subject can be impenetrable to mere mortals.

And, just to add to it, I really don't like the idea of saying that aRGB has more colors than sRGB. Because in fact, it has the exact same number of RGB combinations. And each RGB combination is a unique color value, mathematically speaking.  When you start saying that some values are so close, as to be the same color, it really confuses the whole concept by bringing perception into muddy the idea.

I would suggest maybe you start with something like this:  Assume two color spaces, space "A" and space "B" where each color space consists of 4 colors.  Color space "A" has it's most saturated colors a little bit less saturated than the most saturated colors in color space "B"...

But both spaces contain ONLY 4 colors.  This is something everyone can see without complex 3d graphs.  Extrapolate from there to the spaces with millions of colors.

To me, sorry to say, when you say that Adobe RGB has more colors than sRGB, it's sounds a bit like that fellow who started this whole mess in the first place  

I didn't even worry about color space until the recent Fong blowup and after perusing various websites, I think I have come to the conclusion that Gary Fong got the right answer using the wrong math. I think these two articles explain it better, at least for me:

http://help.smugmug.com/customer/portal/articles/93362
http://www.cambridgeincolour.com/tutorials/sRGB-AdobeRGB1998.htm

For workflow purposes, why put yourself through extra steps by shooting aRGB? You wind up having to convert it to sRGB a lot of times for sharing online/on tablets/etc. and even for some printing services apparently.  So it seems like for most (not all) people, it'd be more time-efficient to shoot JPG + RAW and have the JPG on sRGB.  That way you can share online/on an iPad/etc. without any more steps, and for the few photos that need printing, you can just go back to the RAW and tinker with it in aRGB or ProPhoto or whatever your destination printer uses.

This won't work for some people but I think it would work for most people. Y'all can argue about color management in an ideal world, but we don't live in such an ideal world yet, so shooting sRGB JPG + RAW seems like the most time-efficient way to deal with current realities, yes?

[Tony Jay]

I didn't even worry about color space until the recent Fong blowup and after perusing various websites, I think I have come to the conclusion that Gary Fong got the right answer using the wrong math. I think these two articles explain it better, at least for me:

http://help.smugmug.com/customer/portal/articles/93362
http://www.cambridgeincolour.com/tutorials/sRGB-AdobeRGB1998.htm

For workflow purposes, why put yourself through extra steps by shooting aRGB? You wind up having to convert it to sRGB a lot of times for sharing online/on tablets/etc. and even for some printing services apparently.  So it seems like for most (not all) people, it'd be more time-efficient to shoot JPG + RAW and have the JPG on sRGB.  That way you can share online/on an iPad/etc. without any more steps, and for the few photos that need printing, you can just go back to the RAW and tinker with it in aRGB or ProPhoto or whatever your destination printer uses.

This won't work for some people but I think it would work for most people. Y'all can argue about color management in an ideal world, but we don't live in such an ideal world yet, so shooting sRGB JPG + RAW seems like the most time-efficient way to deal with current realities, yes?

You do realise that RAW files cannot have a colourspace assigned to them in camera like a JPEG.
Until a RAW has been demosaiced strictly speaking it doesn't have any colour detail at all just greyscale tone and so, by definition cannot have an assigned colourspace.
If you are using Lightroom or Adobe Camera Raw then they perform this demosaicing. All the image colour data is then interpolated into the working colourspace which happens to be ProPhotoRGB. Not until you convert this RAW file into some other file type - TIFF, PSD, JPEG etc will any colourspace be embedded.

Now there are situations where just shooting JPEG's and assigning them either AdobeRGB or sRGB in camera is a good idea and there are situations where shooting RAW is an excellent idea too.
However, asking the camera to save both a RAW and JPEG whenever you press the shutter, is, nearly all the time, not getting the best of both worlds as far as image quality goes (either the JPEG or the RAW will suffer) and is usually a colossal waste of storage space.
You do not sound, to me, as if you are the sort of photographer with very complicated workflow requirements that are edge-case where it might make sense to shoot both RAW and JPEG together.

Why do I say what I say above: well optimal exposure for a RAW file is usually (often) very different to exposing for a JPEG. JPEG's need to be exposed very similar slide film while RAW files are optimally exposed using a principle called Expose To The Right (ETTR), which, if you are also saving JPEGS will result in a hopelessly overexposed JPEG.
Exposing optimally for a JPEG however leaves an enormous amount of potential tonal information on the floor as far as the RAW file is concerned.
Since RAW files usually 14-bit files (and sometimes more) and JPEG files are 8-bit files trying to pull and push those JPEG's around in the Develop module of Lightroom, or in ACR, will make them fall apart, while RAW files not optimised by ETTR will show excessive noise in the shadows that becomes especially evident as soon as one tries to lift the shadows.
(It is true that recent late-model cameras have a much lower noise-to-signal ratio so the noise issues have abated a bit.)

If some of this has come as a bit of a surprise to you I encourage you to research the relative merits of RAW capture vs using a JPEG only workflow using the excellent resources found on this site in the various articles and tutorials.

I make no comment as to what specific option may be best for you but suggest it may be both redundant and expensive to continue to shoot both RAW and JPEG.

Tony Jay

[Simon J.A. Simpson]

I think trying to talk about the “number of colours” in a given colour space is a complete red herring and will confuse the target audience of Andrew’s proposed video.

Trying to keep it simple…

As I understand it (and I am ready to be corrected on this) a colour space defines a gamut or ‘range’ of colours.  It is, I believe, misleading to talk about a gamut or range as containing a given “number” of colours.  To put it very simply the gamut or range defines the boundaries of what can be represented.

Still keeping it simple my understanding is that bit depth defines the number of differences between colours that can be represented and would, therefore, be colour space agnostic.

I think if I were asked by a neophyte photographer “Does sRGB have more colours than Adobe RGB ?”, my answer would be “No, but Adobe RGB can contain a wider range or gamut of colours than sRGB.  It is not about the ‘number’ of colours but about the range that can be represented [here a picture of the two colour spaces would depict this very well and aid understanding].”.   I would go on to emphasise that in terms of colour spaces it is not the ‘number’ of colours than can be represented but the range or gamut that is important.  I think this is all someone who asks this question would need to know.  If they want to know more there are plenty of excellent books out there – authors Andrew Rodney and Jeff Schewe to name but two.

If it were me I would forget the thermometer analogy and stick to a picture of the two colour spaces (2D or 3D – doesn’t matter).  These show exactly what we are talking about (and no one need die from overheating !). 

[bjanes]

If I can make a suggestion - the best success that I've had in talking to people for whom "gamut" wouldn't be appropriate is to talk about a "range of colors". So e.g., Adobe RGB can represent a greater range of colors than sRGB.

Talking about "number of colors" generally results in total confusion, because it confuses number of bits in a particular file format with color space.

So actually, in an 8-bit JPEG file, then sRGB and Adobe RGB can represent exactly the same number of colors (depending on how you count colors, which another topic on its own, say 16.7 million odd). However:

• The sRGB version of the file can show a smaller range of colors, but with greater granularity (better color resolution)
• The Adobe RGB version of the file can show a greater range of colors, but with lower granularity

In practice the ability of the human eye to perceive subtle differences in color isn't too great, so the likelihood is that you won't actually see any difference from the granularity.

Sandy

Sandy,

I like this explanation. The thermometer analogy can be extended a bit. Take the example of a laboratory thermometer that covers the temperature range of the transitions of water from freezing to boiling (0 to 100 degrees Celsius or 32 to 212 degrees Fahrenheit). Further assume that the length of the thermometers is 30 cm (11.8 inches).The position of the mercury is a measure of the temperature and is a continuous variable. The range of both scales is the same and it is possible to include both scales on the thermometer, side by side. If we quantize the temperature by reading to the nearest degree, the Celsius scale will have 100 gradations and the Fahrenheit scale will have 180 gradations and offer more precision. However, we could mark the Celsius scale in 0.5 degree increments and the Celsius scale would then have more precision. This is analogous to using a bit depth of 8 for sRGB and a depth of 16 for the ProphotoRGB. The analogy breaks down for the color spaces, because the upper bound of the ProPhotoRGB is greater than that of sRGB. This is like extending the length of the thermometer for Celsius to 50 cm and leaving it the same for Fahrenheit. The Celsius thermometer can now cover a greater temperature range. However, in both cases, one can not usefully extend the granularity of the markings beyond the resolution of the observer's perception (analogous to delta E).

Bill

[mrenters]

I think one of the things we need to keep in mind is that we are talking about photography and photographs that will be viewed by humans.  If we as humans can't perceive any differences between two RGB values in a final output image, then practically it doesn't really matter if those two pixels have the same or different RGB values - they look the same to us.  I think this is important to keep in mind, especially if the goal is to make a video for people that are just starting out. High school physics starts out with Newtonian physics and not the theory of relativity because the former is easier to understand and sufficient to explain most of the interactions we are likely to run into.

I woke up this morning and Andrew's question of does a 16 bit sRGB image of Bill's flower have more colours than an 8 bit AdobeRGB version was on my mind. If we define colour as unique RGB value, then it certainly does assuming of course that the image has always been in a 16 bit depth. If we define colour as something I can uniquely discriminate, it seems reasonable to me that if I can't discriminate between all the 8 bit sRGB colours, then making them 16 bit isn't going to change anything.  Yes, there will be more distinct RGB values in the 16 bit image but if I can't see the difference, then it doesn't really matter much, does it?

This is course opens up whether we should be capturing and editing in higher bit depths. I would definitely argue that we need to, even if we can't see any difference because it gives us more latitude to manipulate and adjust the image before it starts to posterize.

Martin