No difference with various rendering intents?

Hello,

I have used a TIF-test-image which is in ProPhoto colour space, loaded it into RT and set the output profile to RT_sRGB. Then I created output images with each of the four rendering intents. Since going from ProPhoto to sRG I expect that there are quite some colours out of gamut. However, all four output images are absolutely identical.

Did I make a mistake or did I misunderstand something?

Hermann-Josef

Hmm, this topic (why isn’t perceptual intent different from relative colorimetric, etc) has been discussed quite a bit in this forum, for example:

There’s also this:

which has a link to a nice Cambridge in Colour tutorial on the topic of rendering intents. Unfortunately the C in C tutorial fails to mention the difference between matrix and LUT profiles.

@patdavid - somewhere on this forum is a proposed list of topics for tutorials. My apologies, I’m not volunteering to write a tutorial on this question of “why doesn’t the intent make any difference”. But it would make a nice tutorial, and I’d be willing to proof read it for technical errors, if someone else wants to write a tutorial on this topic.

2 Likes

Thanks Elle for the links. Reading your page it occurred to me that the ProPhoto-ICC-profile is also “only” a matrix profile, and as such can only provide the colorimetric intent for mathematical reasons. This should explain why all images come out the same. One suggestion from my side would be to disable those intents in the GUI which are not applicable due the nature of the profiles involved. This would avoid confusion.

I will look for the LUT-version of ProPhoto and see, if this makes then a difference in the output. I just found out that the RT-profiles for sRGB and AdobeRGB provide the LUT-versions of these profiles.

Hermann-Josef

When did RT start supplying LUT versions of sRGB and AdobeRGB? I looked at the RT profiles (and also the profile-making code) not too long ago and didn’t see any LUT profiles, but maybe I overlooked them.

You could experiment with using the color.org sRGB LUT profiles to squeeze your ProPhotoRGB colors down to fit in sRGB. I’d be curious to see how this works out.

Hmm, actually in your case, if I read your first post correctly, you are converting from ProPhoto (the source profile) to sRGB (the destination profile). So what you need is a LUT destination profile, hence a LUT version of sRGB.

@Elle I read the rawpedia page on colour management. There are descriptions of the profiles supplied and they refer to the number of LUT points. Am I misinterpreting this?

Yes, I want to convert from ProPhoto to sRGB, just for curiosity to see how the intents work. I would have thought that the RT-sRGB-profile would allow me to do this. But probably I am wrong???

I looked at color.org but could not find the LUT-versions of sRGB or ProPhoto there. Do you have a link to them?

Hermann-Josef

If I got this right, the RT_* and Rec2020 profiles are matrix profiles with 4096 point TRC LUTs.

The ACES TRC seems to have no LUT and a gamma of 1.0.
The DCI-P3 profiles have a TRC with no LUT and a reported gamma of 2.6, though I’m baffled how that curve shape can result from a gamma of 2.6.

@Morgan_Hardwood Thanks for the clarification. So I was misunderstanding this. If I cannot find the appropriate LUT-based ICC-profiles I will have to make one. Graeme Gill gave me a brief recipe on how to accomplish this with Argyll.

Hermann-Josef

Go here and accept the terms or else you can’t go any further:

Then click on “color encoding profiles” where it says

Profiles

A range of ICC profiles are provided, including:

  • Colour encoding profiles - for RGB and XYZ encodings

Then under “Colour encoding profiles”

click on sRGB profiles" for the sRGB LUT profiles, and on “ROMM RGB” for the ProPhoto LUT profile.

Here are the direct links:

For sRGB you want the “Preference” version.

It’s a good idea to read the documenation for how the ICC intends these profiles to be used.

That would be cool if you could share how to do this.

@Elle Thanks a lot for the links and the instructions!

Here is what Graeme Gill wrote:

Basic outline using ArgyllCMS: create a target sample set using targen,
create measurement samples using fakeread, create profile using
colprof with a source gamut selected.

I have not yet tried to do it.

Hermann-Josef

I’m very interested in how well these profiles work for your particular image, if you wouldn’t mind posting the before and after, assuming of course that the image is of the type that you are comfortable posting publicly.

I tried using those profiles a long time ago to go from ProPhotoRGB to sRGB, and the resulting image didn’t look so good. My sample image was very saturated from the camera (a bright red flower), and I had added extra saturation on top because I needed a sample image for a tutorial (this tutorial: Clipped colors from ICC profile conversions).

Results from using the color.org profiles are not in the tutorial linked to in the previous paragraph, because that’s not what the tutorial is about. But the problem with any LUT profile is that as Graeme Gill noted, you need a source and destination set of colors, and the colors in my very saturated red flower fell outside the source set of colors used to make the color.org LUT profiles.

In case you are curious about the source colors in the color.org profiles, quoting the color.org documentation for the sRGB preference profile, it’s the:

Perceptual Medium Reference Gamut defined in the ICC v4 specification as an intermediate rendering target.

I’ve been told that the PMRG actually works pretty well for a lot of images, so I’m not saying “that stuff isn’t any good”, I’m just saying “it probably won’t work so well if the source image is really, really colorful, with colors outside the PMRG”.

@Elle Having tried it quickly, I get the same result: No difference. I used the sRGB2014.icc profile, since the sRGB_v4_ICC_preference.icc did not show up in RT. Why, I do not know.

For comparing the resulting images I had used imageJ and subtracted the images. I hope this is correct…

Hermann-Josef

Well, it has to be the preference profile or at least a LUT profile of some sort. I’m not sure which profile the sRGB2014.icc profile is, but if that’s not a LUT profile, then it will give “no difference”.

I saw the preference profile show up in the RT drop-down when first I looked, then later it vanished. So maybe RT doesn’t support choosing a LUT output profile. @Morgan_Hardwood ? @heckflosse ?

Anyway, here’s a screenshot from GIMP of my very saturated red flower before and after, using “perceptual intent”:
preference

Keep in mind, that’s a very saturated and bright red flower, totally unprintable in its “before” condition.

No, scratch that screenshot. it’s wrong, there was a problem with how GIMP treated the TRCs. Sigh. I like the above screenshot more than the actual results - look inside the faint white lines for areas that have been through the preference profile, leastways I hope this is right:

Also, if your monitor profile’s color gamut is smaller than sRGB, then you might not see any “before/after” difference.

As a very important aside, I’ve never used these LUT profiles for actually converting between color spaces (other than the one test long ago), and only skimmed through the documentation, again a long time ago. But I think for best results (which I didn’t do with the above screenshot), first you convert from the matrix ProPhoto profile, to the LUT ROMM profile, and then to the LUT sRGB profile. But this is just a guess. Results following this procedure for the red flower image actually are pretty good - I used perceptual intent for both conversions to try this approach (sorry, no screenshot).

1 Like

@Elle Thanks for your example. I will have to look into this in much more detail, but it will take time. I will also try to create a LUT-based sRGB profile with Argyll. Once I got there I will let you know.

Hermann-Josef

1 Like