Color management myths and misinformation video

[digitaldog]

Have you made a video yet Andrew?

Nope, but I'm working on it. The test file I just referenced is for those who want to forgo having to shoot something and just want to see the effect on their printer using a wide gamut source working space and sRGB or something smaller out to print.

[GWGill]

And no, delta E cubed would NOT be a gamut measure, unless you believe that luminance is a component of gamut. (Is RGB 10,10,10 a different color to RGB 20,20,20?)

This may be news to you, but color spaces are 3 dimensional. Therefore their gamut is a volume.

You may have been mislead by the tendency of many people only to show chromaticity graphics - quite understandable due to the difficulty of presenting 3 dimensional data in a 2d media.
 

All of which I think makes the point - if a correct measure doesn't immediately come to the mind of the author of ArgyllCMS, then a generally accepted measure doesn't exist outside of deep geek world.

Ah - I see - you're a troll.

[aaronchan]

Hi Andrew,

Do you have any access to a "sRGB printer"? you know what I mean by that.
if you can have access to that so you can demonstrate and explain why noritsu or fuji controller are/aren't doing the transformation correctly or not.

aaron

[sandymc]

Ah - I see - you're a troll.

Once you've been active on the forum a while longer, you might find out different.

Yes, my use of luminosity was technically incorrect, but this thread started as a way to communicate with people that don't know much at all about color management. If they have ever seen a plot of Adobe RGB or sRGB, it will have been a 2-D xy plot. So an area plot.

Yes, technically gamuts (at least in LAB, etc) are 3D. But good luck trying to address that to this audience.

And even in a volume space, Delta E cubed still wouldn't be a useful way of measuring gamut. Delta E is at its simplest form the length of a line in 3-space (aka in it's simple form, a square root of a sum of squares) - volume calculations don't need or want squares and square roots. In more complex forms that adjust for perceptual uniformity, cubing Delta-E is entirely meaningless.

Sandy

[Bart_van_der_Wolf]

Delta E is at its simplest form the length of a line in 3-space (aka in it's simple form, a square root of a sum of squares) - volume calculations don't need or want squares and square roots.

Hi Sandy,

Don't mistake the integer coordinates (e.g. 0 ... 255) you know from RGB coordinate systems with their origin, which are floating point approximations of a continuous colorspace, usually in XYZ coordinates or their conversion to L*a*b* space.

By changing the colorspace, straight line relationships in one space can be non-linear relationships in another space. Therefore cubed delta-E can be e.g. elliptical volumes in another space (the mathematical cube operation in floating point notation can result in extremely smooth shapes, not just "cubes").

It's all just a bit difficult to show in 2-D display or print, therefore interactive VRML display of 3-D gamuts are commonly used.

Cheers,
Bart

[fdisilvestro]

this thread started as a way to communicate with people that don't know much at all about color management. If they have ever seen a plot of Adobe RGB or sRGB, it will have been a 2-D xy plot. So an area plot.

Yes, technically gamuts (at least in LAB, etc) are 3D. But good luck trying to address that to this audience.

I agree that it is not an easy task, but I think that 3D is the way to go. Not only LAB, in RGB or XYZ you also have three "dimensions".

It seems to me that we try to associate dimensions with physical equivalents, so 2D is area and 3D is volume, but what do you do when you have a n-dimensional problem? Even if we try to simplify, there is a "floor" of complexity that cannot be lowered without introducing errors.

The self proclaimed owner of the truth in the other closed thread just went too far in his "rainbow" simplification (and even seemed that he did not understand the subject anyway)

Regards,

[sandymc]

Hi Sandy,

Don't mistake the integer coordinates (e.g. 0 ... 255) you know from RGB coordinate systems with their origin, which are floating point approximations of a continuous colorspace, usually in XYZ coordinates or their conversion to *L*a*b space.

By changing the colorspace, straight line relationships in one space can be non-linear relationships in another space. Therefore cubed delta-E can be e.g. elliptical volumes in another space (the mathematical cube operation in floating point notation can result in extremely smooth shapes, not just "cubes").

It's all just a bit difficult to show in 2-D display or print, therefore interactive VRML display of 3-D gamuts are commonly used.

Cheers,
Bart

Bart, it doesn't have anything to do with integer or float; if you want to compute the volume of any shape its an integral over all three dimensions. To go mathematical, a triple integral of the constant function over its volume. For a cube, that simplifies to multiplying the lengths of the sides. What it is NOT is the triple integral of the delta E function, which is what is being claimed above.  That makes no sense - it's a mathematical nonsense.

Sandy

[fdisilvestro]

Bart, it doesn't have anything to do with integer or float; if you want to compute the volume of any shape its an integral over all three dimensions. To go mathematical, a triple integral of the constant function over its volume. For a cube, that simplifies to multiplying the lengths of the sides. What it is NOT is the triple integral of the delta E function, which is what is being claimed above.  That makes no sense - it's a mathematical nonsense.

Sandy

I don't know precisely the delta E function, but I would like to know if it is a conservative one. I mean, if I move a specific amount of delta E in one dimension, then another movement in another dimension and one movement in the remaining direction, will I get to the same point regardless of the order of operations and can go back to the point of origin using the same delta E steps? If it is not conservative, then I will have to agree that it makes no sense

[sandymc]

I don't know precisely the delta E function, but I would like to know if it is a conservative one. I mean, if I move a specific amount of delta E in one dimension, then another movement in another dimension and one movement in the remaining direction, will I get to the same point regardless of the order of operations and can go back to the point of origin using the same delta E steps? If it is not conservative, then I will have to agree that it makes no sense

Francisco, there are different Delta E functions, but almost all are a square root of squares (of real numbers). So the function is always positive, hence you can never get back to point of origin.

Sandy

[fdisilvestro]

Francisco, there are different Delta E functions, but almost all are a square root of squares (of real numbers). So the function is always positive, hence you can never get back to point of origin.

Sandy

Sandy, I'm not sure I follow you, I understand that delta E is a "distance" or equivalent. Going back not necessarily requires the distance to be negative, just change the direction.

Maybe I change my question to:
In the L*a*b space, aren't the units in the coordinates related to delta E?

[GWGill]

Francisco, there are different Delta E functions, but almost all are a square root of squares (of real numbers). So the function is always positive, hence you can never get back to point of origin.

The many refinements of DeltaE functions used for the specific purpose of calculating delta E's have little or nothing to do with measuring gamut. The point is that a perceptually uniform colorspace strives to have coordinate units that are one delta E. So by definition volume in such a space is delta E cubed.

[GWGill]

In the L*a*b space, aren't the units in the coordinates related to delta E?

They are CIE76 delta E.

DIN99 space could also be used as an improved perceptual space in which to measure gamut.

[sandymc]

Sandy, I'm not sure I follow you, I understand that delta E is a "distance" or equivalent. Going back not necessarily requires the distance to be negative, just change the direction.

Maybe I change my question to:
In the L*a*b space, aren't the units in the coordinates related to delta E?

Francisco,  related yes. But e.g., the absolute value of x is related to x. That doesn't mean that you can interchange them. 

Sandy

[sandymc]

The point is that a perceptually uniform colorspace strives to have coordinate units that are one delta E. So by definition volume in such a space is delta E cubed.

Noooooooo!

A perceptually uniform colorspace tries to have delta E values that correspond to color differences as perceived human vision. So if the perceived difference between color A and color B and between color C and color D is the same, then the delta E should be the same. There is a belief (not everyone agrees) that a delta E of 1 corresponds to the smallest perceivable color difference.

None of that means that delta E is a coordinate unit of anything.

Again, this time with meaning - delta E is a derived measure of color difference. That's all. It is not a measure of gamut. You cannot sensibly plot a gamut in delta E units and you cannot sensibly measure the volume of a gamut in delta E cube units.

Sandy

[GWGill]

Again, this time with meaning - delta E is a derived measure of color difference. That's all. It is not a measure of gamut. You cannot sensibly plot a gamut in delta E units and you cannot sensibly measure the volume of a gamut in delta E cube units.

Exactly, DE76 is the straight line distance between any two point in L*a*b* space, which can only be the case if units of L*a*b* space are DE76.

There's nothing complicated going on there, it's exactly like plotting real world objects against 3 orthogonal axes. If you want to compute the distance between any two points in such a coordinate system in (say) meters using straight line between the points, then the axes will also have to be measured in meters.

[sandymc]

Exactly, DE76 is the straight line distance between any two point in L*a*b* space, which can only be the case if units of L*a*b* space are DE76.

There's nothing complicated going on there, it's exactly like plotting real world objects against 3 orthogonal axes. If you want to compute the distance between any two points in such a coordinate system in (say) meters using straight line between the points, then the axes will also have to be measured in meters.

You're trying to make a "the lab value is a GPS coordinate, and the delta E is like mileage" analogy?

Well, I guess as an analogy that might be useful, but it's really not like that as a physical measurement. E.g., mileage is an independent physical measure (multiples of a wavelength of specific light these days, I think). Delta E is only defined as difference between lab values, without a physical unit.

But good luck. I will watch the video that results from this thread with interest.

Sandy

[fdisilvestro]

Delta E is only defined as difference between lab values, without a physical unit.

Well, if we agree that color is not a physical property but a perception of our brain, then it cannot be possibly be described by a physical unit. The "volume" in this case is an abstraction.

[GWGill]

You're trying to make a "the lab value is a GPS coordinate, and the delta E is like mileage" analogy?

There is no analogy here, DE76 is the distance between two points in L*a*b* space.

But you'd be much more helpful if you went and looked some of these things up
rather than causing confusion by making assertions off the top of your head.

[sandymc]

There is no analogy here, DE76 is the distance between two points in L*a*b* space.

But you'd be much more helpful if you went and looked some of these things up
rather than causing confusion by making assertions off the top of your head.

Graeme,

So far in this thread you've previously accused me of being troll, and now the above.

I have questioned some of the assertions you've made, but have refrained from any name calling, or questioning your helpfulness, motives for being here, or whatever else.

I'd suggest that you rethink your attitude before the mods rethink it for you. But however the rest of your stay on this forum turns out, you're now on my personal ignore list.

Sandy

[GWGill]

But however the rest of your stay on this forum turns out, you're now on my personal ignore list.

Perfect - you're going on mine. I just pity all those being mislead by your determination to half understand color.