Another thread on Relative vs Absolute Colorimetric intent

[nand]

Preamble: If this thread is posted in the wrong section, please move it - I don't quite know where to best post this.

I've read the articles at http://www.steves-digicams.com/knowl...g-intents.html, https://www.cambridgeincolour.com/tu...conversion.htm and http://en.wikipedia.org/wiki/Color_m...ndering_intent but I still don't quite understand why relative colorimetric intent is preferable to absolute colorimetric (for monitors).

If I'm displaying a photograph or movie file on my monitor (which has a white point slightly off from the ideal 6504K for sRGB), wouldn't I want to use absolute colorimetric mode to ensure that colors display the way they would on an ideal sRGB monitor?

In particular, this quote confuses me: “...using the absolute colorimetry rendering intent would ideally [...] give an exact output of the specified CIELAB values. Perceptually, the colors may appear incorrect, but instrument measurements of the resulting output would match the source.”

If the instrument measurements of the actual colors emitted from the monitor are 100% identical to the source material I'm viewing (photograph, movie frame), then how come colors would appear incorrect?

Wouldn't it look *exactly* the same as if I had viewed the same scene in real life (assuming the camera and display are perfect)?

Furthermore, if an image is tagged as sRGB (6504K white point) and I view it on a monitor with a 5500K white point, wouldn't I want the image to display using the actual, blueish hue that I would get on a perfect sRGB monitor? I wouldn't want images to appear with a 5500K white point because that would not be accurate to the source at all.

Why is it, then, that everywhere I see, relative colorimetric mode is advertised instead of absolute? It seems counter-intuitive to me.

Any explanations would be greatly appreciated.

[Colin Southern]

Preamble: If this thread is posted in the wrong section, please move it - I don't quite know where to best post this.

I've read the articles at http://www.steves-digicams.com/knowl...g-intents.html, https://www.cambridgeincolour.com/tu...conversion.htm and http://en.wikipedia.org/wiki/Color_m...ndering_intent but I still don't quite understand why relative colorimetric intent is preferable to absolute colorimetric (for monitors).

If I'm displaying a photograph or movie file on my monitor (which has a white point slightly off from the ideal 6504K for sRGB), wouldn't I want to use absolute colorimetric mode to ensure that colors display the way they would on an ideal sRGB monitor?

In particular, this quote confuses me: “...using the absolute colorimetry rendering intent would ideally [...] give an exact output of the specified CIELAB values. Perceptually, the colors may appear incorrect, but instrument measurements of the resulting output would match the source.”

If the instrument measurements of the actual colors emitted from the monitor are 100% identical to the source material I'm viewing (photograph, movie frame), then how come colors would appear incorrect?

Wouldn't it look *exactly* the same as if I had viewed the same scene in real life (assuming the camera and display are perfect)?

Furthermore, if an image is tagged as sRGB (6504K white point) and I view it on a monitor with a 5500K white point, wouldn't I want the image to display using the actual, blueish hue that I would get on a perfect sRGB monitor? I wouldn't want images to appear with a 5500K white point because that would not be accurate to the source at all.

Why is it, then, that everywhere I see, relative colorimetric mode is advertised instead of absolute? It seems counter-intuitive to me.

Any explanations would be greatly appreciated.

Hi Nand,

RC will preserve more of the original colours and fit in better with the gamut of the output device (kinda like "when in Rome, do what the Romans do") - AC is more for trying to simulate the output of one type of device, on another device.

In real-world terms, because our eyes adapt to the brightest thing being white, RC generally just works better in practice (in reality it's all I ever use).

If in doubt, just do some simple experiments and see for yourself

[nand]

Hi Nand,

RC will preserve more of the original colours and fit in better with the gamut of the output device (kinda like "when in Rome, do what the Romans do") - AC is more for trying to simulate the output of one type of device, on another device.

In real-world terms, because our eyes adapt to the brightest thing being white, RC generally just works better in practice (in reality it's all I ever use).

If in doubt, just do some simple experiments and see for yourself

I believe I understand now. In other words, due to our brain interpreting the light from everything else around us as being “white”, the correct white looks blue-ish as a result.

So, as a result, if I work in a completely dark room with no other light source than my monitor and view a photograph or movie file in fullscreen, I would notice no color hue at all since my eyes would adapt to the displayed 6504K white as being white (even though my display might be something entirely different)?

[nand]

By the way, I don't want to open a new thread for this, but does anybody know why some ICC profiles have multiple AToB or BToA tables?

What kind of magical monitor displays different colors for the same input values (depending on, what, I don't know, phase of the moon?)?

[Colin Southern]

I believe I understand now. In other words, due to our brain interpreting the light from everything else around us as being “white”, the correct white looks blue-ish as a result.

So, as a result, if I work in a completely dark room with no other light source than my monitor and view a photograph or movie file in fullscreen, I would notice no color hue at all since my eyes would adapt to the displayed 6504K white as being white (even though my display might be something entirely different)?

Yes, pretty much. Sometimes it's a curse, sometimes it's a blessing. I've got a pretty good eye for colour and even I can look at something with sensitive colour requirements (eg skintones) - adjust it by eye to "perfection" and still get it waaaaaaaaaay out of whack.

[Colin Southern]

By the way, I don't want to open a new thread for this, but does anybody know why some ICC profiles have multiple AToB or BToA tables?

One for each rendering intent, and in each direction.

What kind of magical monitor displays different colors for the same input values (depending on, what, I don't know, phase of the moon?)?

I could swear mine does somedays, but it's probably just me

[nand]

What do rendering intents have to do in the context of a device-independent PCS such as XYZ, though?

Or can an ICC profile only have multiple rendering intents when it's a profile with device-dependent input *and* output spaces? eg. a RGB->RGB profile.

I can see no way how it would be possible otherwise.

[Colin Southern]

Hi Nand,

Short answer is "not a lot", but then again, you didn't specify only device independent profiles in your previous question.

Dinner time for me (tail end of a tiring day), so probably the best thing I can do is suggest you grab yourself a copy of Real World Color Management, by Fraser, Murphy, and Bunting - it's the industry-standard text on colour management, and covers these types of questions in chapter 3. Also available in Kindle edition (recommended).

Hope this helps.

[nand]

Thank you for these suggestions. I've actually just gotten an answer from the ICC mailing list on my question, the short version is: [in the context of RGB->XYZ profiles] “there's no clearly defined common gamut boundary, so how to implement the perceptual / saturation modes is implementation-specific”.

A link explaining it: http://www.argyllcms.com/doc/iccgamutmapping.html

I'll grab a copy of that book though, it looks promising.

[PhotomanJohn]

Back to RC and AC. My simple minded understanding is that with AC, the colors that are not within the current gamut are simply changed to fit them into the gamut without adjusting any of the other colors essentially by assigning them the same values as the closest one within the gamut.

PC does a similar thing except that it adjusts colors that are already in the gamut such that there is "room" for the ones being pulled in from the outside so they are not assigned the exact value of another color.

I am told that AC can cause visible banding where colors are gradually changing and other artifacts in some pictures.

This is just what I have been told but don't really know if there is an ounce of truth in it. I would appreciate someone correcting me as needed.

John

[nand]

Back to RC and AC. My simple minded understanding is that with AC, the colors that are not within the current gamut are simply changed to fit them into the gamut without adjusting any of the other colors essentially by assigning them the same values as the closest one within the gamut.

Both colorimetric modes work like this, that includes absolute and relative colorimetric.

PC does a similar thing except that it adjusts colors that are already in the gamut such that there is "room" for the ones being pulled in from the outside so they are not assigned the exact value of another color.

Perceptual mode works like this, yes - but it is not a colorimetric mode.

I am told that AC can cause visible banding where colors are gradually changing and other artifacts in some pictures.

Perceptual mode can introduce banding - colorimetric mode can introduce clipping. The relative colorimetric mode might also introduce minor banding due to the white point transformation, absolute colorimetric mode should introduce no banding at all.