sRGB vs. Adobe RGB: New color management stand up comic

[digitaldog]

Personally, to try quickly a browser I like the Petapixel example is it leaves no doubt thanks to the funky profile used.
http://petapixel.com/2012/06/25/is-your-browser-color-managed/
This one is more theoretical and complete:
http://cameratico.com/tools/web-browser-color-management-test/

Garry's page certainly isn't as pretty but what I like is he explains how all the issues in detail and how to set them up, also what is assumed for untagged data.

[supercurio]

We're now doing tests on our mobile devices and uploads to Facebook.  I will do a video on YouTube to show all of these results.

Only Firefox/Android supports color management so far, last time I checked it was requiring enabling a hidden setting manually tho it might have changed since.

On this topic, please don't make the typical mistake to say "smartphones and tablets displays are sRGB" like Will did in his video on your channel.
Mobile devices right now both lack any sort of color management while they're the ones needing it the most due to incredible variety between displays.

From cheap or because much tuned for power efficiency panels with crazy small gamuts to widest on the industry right now (using both AMOLED or IPS LCD)
Add that to far too often gamma curves, mobile devices can't be resumed as "sRGB", more like a mess instead  

[digitaldog]

On this topic, please don't make the typical mistake to say "smartphones and tablets displays are sRGB" like Will did in his video on your channel.

But aren’t the pixels farther apart, that's what the faux Coloratti said in that video?  

[Schewe]

But aren’t the pixels farther apart, that's what the faux Coloratti said in that video?  

Ya know, somebody should let Liz know what's he's doing…I'll try to send her an email (she still likes me :~)

[supercurio]

But aren’t the pixels farther apart, that's what the faux Coloratti said in that video?  

I wouldn't blame people too much not knowing mobile in details.
Today it's a domain very few investigated.
Those who do don't necessarily understand all the specifics of this platform, ending sometimes into some large mistakes in their analysis and evaluations.

That's why I'm available for any question on the subject and offer help and tools to reviewers.

[digitaldog]

Ya know, somebody should let Liz know what's he's doing…I'll try to send her an email (she still likes me :~)

Go for it, (I didn't want to be a tattle-tail  )

[digitaldog]

I wouldn't blame people too much not knowing mobile in details.

Agreed, I know next to nothing about em.
FWIW, pretty sure he was referring to non-mobile devices.

[bjanes]

It seems he's deleted most of the critical comments.

My comment that I probably should have just passed on:

and his response:

I'm still not sure where those numbers — 1.3M and 900K — come from. Considering an 8bit image has 1.7M possible values, it suggests that I should be able to find 300K AdobeRGB triplets that are somehow not in AdobeRGB and 700K sRGB values (almost half of the available values) that aren't sRGB colors. It makes zero sense. I asked if he would provide just one of the 300K adobeRGB triplets, but he didn't.

Those values represent the gamut volume in cubic DeltaE values. 1 DeltaE is a just perceptible difference in color, which has 3 dimensions, hence the cubic value. Norman Koren briefly describes how it is calculated and points out that it can be visualized by the 3D gamut plots provided by Colorthink or his own Gamutvision program. Bruce Lindbloom lists the gamut volume for various color spaces as follows: L*a*b, 2,381,085; Adobe RGB, 1,208,631, sRGB, 832,658; and ProPhotoRGB 2,879,058. The L*a*b efficiencies (the percent of the entire Lab Gamut [i.e. all colors visible to the eye] that the working space encompasses) are 97.0%, 100%, 100%  and 91.2% respectively. This article from RIT (the link is to a PDF) describes the calculations in more detail.

The L*a*b gamut efficiency is less than 100% with integer encoding. Some of the values in ProPhotoRGB are outside the gamut of human vision and do not represent real colors, which are visible by definition.

Bill

[MarkM]

Thanks Bill,

I finally recognized those values from the volume given by ColorThink. The Norman Koren method of calculating gamut volume is new to me - thanks for the link. A while back I tried to figure out why Bruce Lindbloom reported lower volumes for standard spaces than ColorThink. This is probably the answer. In case you're interested, I used a simple Monty Carlo method and came in more or less exact agreement with Bruce's numbers: http://www.photo-mark.com/notes/2011/jun/29/calculating-color-space-volumes/

I'm not sure I would go as far as to equate the number of perceivably unique colors in a file with the "number of colors" (and I'm pretty sure that's not what Gary meant) unless it's clear we're talking about the perceivable values rather than the measured values. Maybe I'm just a pixel peeper, but it seems that unique values are pretty important for the reasons Mr. or Ms. Eyeball listed in his/her caveats above. As far as I'm concerned a 24-bit file has 24 bits of information even if I can't visually distinguish them.

[digitaldog]

I'm not sure I would go as far as to equate the number of perceivably unique colors in a file with the "number of colors" (and I'm pretty sure that's not what Gary meant) unless it's clear we're talking about the perceivable values rather than the measured values.

He's not talking about any of that. Here's exactly what I heard from Gray in terms of 'number of colors' after showing two spectrum's and specifying them as sRGB and Adobe RGB (1998)

0.57.…And the colors at either side of the rainbow are the same (not true). What the difference is between Adobe RGB and sRGB is there are more colors in-between.

3:19: People will tell you Adobe RGB is better because it has more color information, that’s true. But the color information in between this is wider. It doesn’t mean it has less colors. It means the values between the colors are more points of data. And really, honestly you can’t see them visually to the naked eye.

4:04 (describing the ColorChecker SG)

…There’s just a whole zillon different color checkers of colors…

[MarkM]

The L*a*b efficiencies (the percent of the entire Lab Gamut [i.e. all colors visible to the eye]

I don't understand how you can have a volume of colors visible to the human eye. If you plot it in xyY and pick any point on the surface, what happens if you make it brighter (i.e. extend directly up the Y axis)? It's still within the spectral limits on the xy axis, so it should still be visible to the human eye, just brighter, yet outside the volume.

I found the plots here: http://www.brucelindbloom.com/LabGamutDisplay.html And they look more like possible surface colors, not all visible colors. In other words, the volume represents all the possible surface colors under a given illuminant (the limit being that the surface can't reflect more power of a particular wavelength than is incident upon it).  The gamut plot in xyY on that page looks exactly like the MacAdam Color Solid, which is a plot of object color space, not human vision. Maybe a distinction without a difference…

[fdisilvestro]

Firefox is fully color management capable in windows and Mac. The issue is that it is not enabled by default and the setting is not straightforward.

In the following link the instructions:

http://cameratico.com/guides/firefox-color-management/

Regards

[digitaldog]

http://www.colorwiki.com/wiki/Color_Management_Myths_26-28#Myth_28:_The_PowerBook_G4_displays_16.7_million_colors_.28or_any_display.2C_for_that_matter.29

The PowerBook G4 displays 16.7 million colors (or any display, for that matter)
This is not true. Don't confuse RGB number combinations with the number of perceivable colors.
I can send 16.7 million different RGB NUMBER combinations to a PowerBook display (3 channels with 8 bits per channel) but it will only display 518,733 different colors. This means that 16,258,483 of the RGB numbers are basically "wasted". Another way of looking at this is to say that the entire gamut of 518,733 colors is chopped into 16.7 million separately addressable "chunks". Problem is, the difference between each of these chunks is smaller than is perceivable by humans. So if you glom chunks together until each blob is just barely perceivably different than the next, you'll end up with 518,733 of them.
That explanation is a bit of a stretch but sometimes it helps to break these things down to understand them. (pun intended)
This confusion is another example of the difference between RGB and CMYK values and actual colors.
Another example of this is with CMYK devices. I can send 100,000,000 CMYK values to print on newsprint (100x100x100x100). Does that mean I'm going to get that many actual colors? No, of course not. If I send those CMYK values to a sheet-fed press on glossy coated paper will I get that many colors? No, but I'll get more than I did from newsprint. I'd probably see even more from an inkjet. While I can address the colors on a press using CMYK combinations, each CMYK combination will not produce a unique color.
Thanks for reading,
Steve Upton


[MarkM]

Adobe RGB is better because it has more color information

That one is especially grating because, as we know, information is very often expressed in bits. But mention bit depth in this context and you're an example of someone who knows nothing about the subject.

[supercurio]

http://www.colorwiki.com/wiki/Color_Management_Myths_26-28#Myth_28:_The_PowerBook_G4_displays_16.7_million_colors_.28or_any_display.2C_for_that_matter.29

This explanation is so much of a stretch that I don't see what it means.
Is it based on the assumption that people don't see difference lower than 3 Delta E?

A bit too much of a generalization, when banding is so visible with non-dithered 8bit per channel.

[digitaldog]

Is it based on the assumption that people don't see difference lower than 3 Delta E?

Probably a dE of 1 I suspect.

[bjanes]

Thanks Bill,

I finally recognized those values from the volume given by ColorThink. The Norman Koren method of calculating gamut volume is new to me - thanks for the link. A while back I tried to figure out why Bruce Lindbloom reported lower volumes for standard spaces than ColorThink. This is probably the answer. In case you're interested, I used a simple Monty Carlo method and came in more or less exact agreement with Bruce's numbers: http://www.photo-mark.com/notes/2011/jun/29/calculating-color-space-volumes/

Mark,

I looked at your Monte Carlo method and found it ingenious. Somewhat off topic remarks are the most valuable parts of this thread.

Bill

[digitaldog]

Somewhat off topic remarks are the most valuable parts of this thread.

Indeed! The very reason not to close it, folks that suggest moving on can just do so.

[MarkM]

Thanks Bill, I wish I could take more credit — just applying an old method.

I agree, it would be odd to close a thread just because people are participating.

[Schewe]

Ya know, somebody should let Liz know what's he's doing…I'll try to send her an email (she still likes me :~)

I did…not bad to get it from both of us...