Luminous Landscape Forum
Raw & Post Processing, Printing => Colour Management => Topic started by: Peter_DL on January 12, 2011, 01:38:03 pm
-
Hi,
This subject was repeatedly discussed in the past,
in particular with regard to ProPhoto RGB and the conversion to smaller target / matrix spaces,
however it could easily be that I missed the final conclusion.
Is it possible now to facilitate any 'smooth' perceptual gamut compression,
or are we still with RelCol ?
Appreciate your comments.
& Best regards, Peter
--
-
It seems to be possible with version 4 profiles
There is a version 4 sRGB profile here (http://www.color.org/srgbprofiles.xalter)
-
It seems to be possible with version 4 profiles
There is a version 4 sRGB profile here (http://www.color.org/srgbprofiles.xalter)
Trying it back and forth, it does not seem to work for the conversion ProPhoto RGB to sRGB.
No merits of this version 4 profile are observed compared to ordinary sRGB IEC61966-2.1.
Thanks for the link though.
Am I doing something wrong ?
Example for reference:
> the left image was left in ProPhoto RGB
> the right image was converted RelCol to ordinary sRGB
It’s well visible on (my) screen that some image details are getting lost upon conversion.
By no means of the version 4 sRGB profile and its Perceptual RI was it possible to prevent this.
Hope it can be seen with the attached screenshot:
for the purpose of clipping it in here, the screenshot was pasted in a new file, monitor profile was assigned and then converted to ProPhoto RGB again. So the attached file is in ProPhoto RGB.
Peter
--
-
In my experience there is a difference when you use the version 4 profile.
Following there is an example with a test picture
The first image is the original in ProphotoRGB. You can appreciate that there are no clipped channels in the histogram
The second image is the conversion to the standard sRGB version 2. It shows clipping especially in the red channel. It doesn´t matter if you choose Perceptual or Relative Colorimetric, as probably everybody agrees in this matrix conversions.
The third image is the conversion to the sRGB version 4 using Relative Colorimetric intent. It is basically similar to the conversion to sRGB version 2, which was expected.
The last image is the conversion to sRGB version 4 using Perceptual Intent. In the histogram it shows a noticeable difference with the previous one (the Rel Col conversion). Even if there is still some clipping, it is less than the Rel Col conversion. That´s why I think that perceptual intent is possible.
-
For this to work as designed, both profiles have to be V4 (keep in mind that by the time the destination profile comes into play, its being handed - usually- Lab).
-
Do you know if there are v4 profiles for ProPhotoRGB and AdobeRGB available?
In the www.color.org website I could find only sRGB v4
-
Do you know if there are v4 profiles for ProPhotoRGB and AdobeRGB available?
Not that I’m aware of.
-
Do you know if there are v4 profiles for ProPhotoRGB (...) available?
ISO22028-2_ROMM-RGB
http://color.org/prmg_gamutwarning.xalter
-
ISO22028-2_ROMM-RGB
http://color.org/prmg_gamutwarning.xalter
So how precisely would these spaces have to be used in order to 'perceptualize' the conversion from ProPhoto RGB to sRGB ?
Tried it - but failed again.
Thanks for your comment.
Peter
--
-
So how precisely would these spaces have to be used in order to 'perceptualize' the conversion from ProPhoto RGB to sRGB ?
Tried it - but failed again.
so maybe first convert perceptual to the PRMG_RGB profile and then perceptual to sRGB V4.
Or try to convert first perceptual to PhotogamutRGB and then relcol to sRGB (regular sRGB).
http://photogamut.org/E_ICC_profile.html
The latter should be a better workflow anyway as the V4 sRGB profile often introduces a color shift.
-
Or try to convert first perceptual to PhotogamutRGB and then relcol to sRGB (regular sRGB).
http://photogamut.org/E_ICC_profile.html
There is indeed a slight but perceivable competitive edge with PhotoGamut RGB as the intermediate space.
Some more image details seem to be maintained.
Peter
--
left image: ProPhoto RGB --> relcol to regular sRGB
right image: ProPhoto RGB --> perceptual to PhotoGamut RGB --> relcol to regular sRGB
-
There is indeed a slight but perceivable competitive edge with PhotoGamut RGB as the intermediate space.
Some more image details seem to be maintained.
the question here is whether or not the colours of this image are clipping in your monitor profile.
ProPhotoRGB is assigned to the file attached but obviously it's a screenshot and therefore the colours are limited to the gamut of your monitor profile anyway...
-
the question here is whether or not the colours of this image are clipping in your monitor profile.
ProPhotoRGB is assigned to the file attached but obviously it's a screenshot and therefore the colours are limited to the gamut of your monitor profile anyway...
Sure - however, seeing the loss of color saturation and image details on my screen (particularly with plain RelCol from ProPhoto RGB to sRGB) implies that the monitor gamut stretches beyond sRGB for the relevant colors here; which can be shown separately by a Customized Proof setup and out-of-gamut marks. Alternatively such visual comparison can be done while having the "Desaturate Monitor Colors" option enabled in the Color Settings. Problem is that such image degradation and loss of details is hard to quantify otherwise than by a visual comparison.
Viewing the above image in a non-color-managed web browser, there should be a slight difference visible regarding the image details of the upper left petal (left vs. right image). Viewing the above image in a color-managed environment such as Photoshop with ProPhoto RGB assigned will require that your monitor gamut stretches beyond sRGB as well for the relevant colors.
However, I agree that this is all a 'construct' to share what I see more clearly on my screen:
PhotoGamut RGB provides some sort of perceptual gamut compression,
whereas it didn’t work for me with the version 4 profiles mentioned initially.
Anyway thanks for your comment.
Peter
--
-
So how precisely would these spaces have to be used in order to 'perceptualize' the conversion from ProPhoto RGB to sRGB ?
Tried it - but failed again.
Thanks for your comment.
Peter
--
The first step of using a Ver 4 profile similar to ProPhotoRGB would be to get the image into that profile. ACR does not render into such a profile, but does render into ProPhoto RGB. To get the image into the Ver 4 profile create a scene referred image using Photoshop and ACR as explained in this ICC paper (http://www.color.org/scene-referred.xalter). To obtain scene referred data one must set the tone curve to linear, but some adjustments may be used as explained here (http://www.color.org/scene_analysis_and_rendering.xalter). One can then convert to a wide gamut v4 profile such as ISO22028-3_RIMM-RGB-exCR.icc or linear_RIMM-RGB_v4.icc. From there, one could convert to a Ver 4 output profile using the methods outlined in paragraph 11 of the paper, using either option A or option B. Current perceptual rendering is not very smart, since it compresses the gamut of an image even if the image contains no out of gamut colors in the destination space.
However, rendering from scene referred to output referred is not a simple process as explained by an excellent paper by Karl Lang (http://wwwimages.adobe.com/www.adobe.com/products/photoshop/family/prophotographer/pdfs/pscs3_renderprint.pdf). Visual editing of wide gamut images is limited by the gamut of the monitor. The better monitors cover AdobeRGB and this is sufficient for many images. The newer ink jets have a gamut that exceeds aRGB for some colors (mainly saturated colors at a relatively low luminance), whereas, due to a gamut mismatch, aRGB has a wider gamut at higher luminances. The human perceptual system is not that sensitive to saturation at low luminances and I am not sure how much difference this expanded gamut at low luminances is perceptible in actual prints. Perhaps Jeff Schewe or Nick Raines can comment.
Regards,
Bill
-
One way to get a ver 4 profile similar to ProPhotoRGB would be to create a scene referred image using Photoshop and ACR as explained in this ICC paper (http://www.color.org/scene-referred.xalter).
another way is to simply download the V4 version of ROMM-RGB: http://color.org/prmg_gamutwarning.xalter / http://color.org/profiles/ISO22028-2_ROMM-RGB.icc
-
another way is to simply download the V4 version of ROMM-RGB: http://color.org/prmg_gamutwarning.xalter / http://color.org/profiles/ISO22028-2_ROMM-RGB.icc
That is obvious, but one must get the image from ProPhotoRGB into the ISO22028-2_ROMM-RGB.icc. Are you merely suggesting assigning the ProPhotoRGB image to the Ver 4 ISO22028-2_ROMM-RGB.icc? Does that work? I have revised my post to clear up the misunderstanding.
Regards,
Bill
-
Are you merely suggesting assigning the ProPhotoRGB image to the Ver 4 ISO22028-2_ROMM-RGB.icc? Does that work?
IMO, yes, it does work (there is a slight difference in highest saturated blues and magenta, but negligible... IMO). You could also convert relcol + BPC (literally "lossless" in 16bit).
Of course, with a decent RAW software you could also render directly into ISO22028-2_ROMM-RGB.
The OP referred to a perceptual conversion to sRGB... so for this particular purpose I think converting relcol + BPC from ProPhoto to ISO22028-2_ROMM-RGB as an "intermediate" color space should be fine in order to utilize the V4 sRGB profile.
Then again, me personally I would use the said PhotogamutRGB profile (or maybe the PRMG_RGB profile) as intermediate color space for such a purpose anyway as the V4 sRGB profile produces color shifts (warm tones turn blueish/greenish).
-
The OP referred to a perceptual conversion to sRGB... so for this particular purpose I think converting relcol + BPC from ProPhoto to ISO22028-2_ROMM-RGB as an "intermediate" color space should be fine in order to utilize the V4 sRGB profile.
Starting with ProPhoto RGB
-> RelCol + BPC to ISO_22028-2_ROMM-RGB
-> Perceptual + BPC to the PRMG_RGB profile (or sRGB_v4)
-> RelCol or Perceptual + BPC to regular sRGB
… the problematic step seems to be the final conversion to regular sRGB.
At the end the results seem to be undistinguishable from ProPhoto RGB -> RelCol to sRGB.
Peter
--
-
-> RelCol or Perceptual + BPC to regular sRGB
perceptual to regular sRGB does not exist. The regular sRGB profile does not contain a perceptual table so if you set perceptual in the conversion settings it's actually relcol. Also BPC is actually "included" in perceptual (but not needed with matrix profiles on both ends going from black L*0 to white L*100 anyway).
Just a note...
… the problematic step seems to be the final conversion to regular sRGB.
At the end the results seem to be undistinguishable from ProPhoto RGB -> RelCol to sRGB.
this is why I would suggest to use PhotogamutRGB as intermediate color space (which should be fine for most purposes). One of the design goals of this colour space was to achieve a visual similarity to sRGB (while actually being much larger than sRGB) so that you could use the profile also in non colour managed workflows.
Basically: avoid bright high saturated colours ;-)
-
perceptual to regular sRGB does not exist. The regular sRGB profile does not contain a perceptual table so if you set perceptual in the conversion settings it's actually relcol. ...Just a note...
That's commonly understood, however, with these profiles i.e. PRMG_RGB (or sRGB_v4) being used as the source space, we just did not want to exclude 'anything' in the course of testing. If only the target space needs to include a perceptual table, which is what conventional wisdom suggests, I'm wondering why we ever started to convert to ISO22028-2_ROMM-RGB first, before targeting the smaller PRMG_RGB (or sRGB_v4) spaces. Anyway.
Peter
--
-
That's commonly understood, however, with these profiles i.e. PRMG_RGB (or sRGB_v4) being used as the source space, we just did not want to exclude 'anything' in the course of testing. If only the target space needs to include a perceptual table, which is what conventional wisdom suggests, I'm wondering why we ever started to convert to ISO22028-2_ROMM-RGB first, before targeting the smaller PRMG_RGB (or sRGB_v4) spaces. Anyway.
ah, okay, I got it!
Well, I think both profiles must provide a perceptual table... but at least the target profile.
What happens when you convert perceptual from ISO22028-2_ROMM to Lab and then relcol+BPC to sRGB?
-
ah, okay, I got it!
Well, I think both profiles must provide a perceptual table... but at least the target profile.
What happens when you convert perceptual from ISO22028-2_ROMM to Lab and then relcol+BPC to sRGB?
ISO22028-2_ROMM-RGB
-> RelCol or Perceptual to Lab
… makes no difference, it’s apparently the same.
I’m hearing you regarding PhotoGamut RGB,
whereas however all these ICC (v4, whatsoever) profile variants seem to be a dead end in the context of this thread - unless I’m still missing something.
Peter
--
-
I’m hearing you regarding PhotoGamut RGB,
whereas however all these ICC (v4, whatsoever) profile variants seem to be a dead end in the context of this thread - unless I’m still missing something.
I don't know about the state of the development and implementation of V4 profiles. I don't use those profiles; actually I do not even use matrix profiles (other than my monitor profile and sRGB for web purposes... but for 600x450px small images on the web I don't care about clipping).
Photogamut works like an oversized printer profile (actually it is a printer profile, technically speaking)... this is why perceptual is available here (and actually works really well). Regarding it's original target (encompassing all printers) it's a bit dated (today some printers provide larger gamuts)... but as an intermediate color space to go to sRGB or similar purposes it's still very useful.
-
... Current perceptual rendering is not very smart, since it compresses the gamut of an image even if the image contains no out of gamut colors in the destination space.
However, rendering from scene referred to output referred is not a simple process as explained by an excellent paper by Karl Lang (http://wwwimages.adobe.com/www.adobe.com/products/photoshop/family/prophotographer/pdfs/pscs3_renderprint.pdf) ...
We still find this astonishing (or sometimes annoying): there are many dedicated tools and sliders to facilitate a perceptual compression along the tonal scale, in order to render tonality from scene to output dynamic range - talking about Brightness & Contrast sliders, a Parametric Tone Curve, or selection-based 'HDR' techniques for tone mapping, etc.
Whereas there seems to be much less support for the orthogonal dimension i.e. color saturation, when we try to facilitate a perceptual compression of scene saturation via Raw / ProPhoto RGB into tiny sRGB - e.g. for a simple Frontier lab print.
Regards,
Peter
--
-
Photogamut works like an oversized printer profile ... this is why perceptual is available here (and actually works really well). Regarding it's original target (encompassing all printers) it's a bit dated (today some printers provide larger gamuts)... but as an intermediate color space to go to sRGB or similar purposes it's still very useful.
Supplementary:
The left image was converted from ProPhoto RGB, relcol to sRGB.
As shown in post #2 there's a perceivable loss of image details.
The right image was converted from ProPhoto RGB, perceptual + bpc to PhotoGamut RGB.
Then, a Local Contrast Enhancement was applied through an inverted Saturation Mask.
Finally the file was converted relcol to sRGB as well.
For the purpose of showing it here, the screenshot was pasted in a new file, monitor profile was assigned and then converted to sRGB again.
Peter
--
-
yep, the PhotogamutRGB version looks much better.
If I would work in ProPhoto I would most likely use PRMG_RGB (or Photogamut) as gamut warning profile (preview off / colour warning on).
Not to edit the files so that they always fit into PRMG_RGB... but just to be warned.
Likewise I use my monitor profile as proof profile (preview off / colour warning on)... so that I can see what I can not see ::)
-
The right image was converted from ProPhoto RGB, perceptual + bpc to PhotoGamut RGB.
Then, a Local Contrast Enhancement was applied through an inverted Saturation Mask.
Finally the file was converted relcol to sRGB as well.
Well, to fill out the side by side, you should have used the same procedure to do a local contrast enhancement on the ProPhoto RGB prior to transforming to sRGB using RelCol. As it is, the second image is indeed better but who's to say it's because of perceptual mapping or the contrast work? As it is now, it's kinda apples/oranges.
-
Well, to fill out the side by side, you should have used the same procedure to do a local contrast enhancement on the ProPhoto RGB prior to transforming to sRGB using RelCol. As it is, the second image is indeed better but who's to say it's because of perceptual mapping or the contrast work? As it is now, it's kinda apples/oranges.
My thoughts exactly.
-
you should have used the same procedure to do a local contrast enhancement on the ProPhoto RGB prior to transforming to sRGB using RelCol.
"you should have ..." :-))
the initial question was whether or not there is an sRGB profile providing perceptual RI. Period. There is none; at least none that works as supposed to. Period.
PhotogamutRGB as intermediate colour space provides perceptual RI and therfore works excellent as a workaround for this particular issue.
Of course you can do a lot of things with curves, saturation, masks and you name it. But that was not the question here.
As it is now, it's kinda apples/oranges.
Not so!
-
"you should have ..." :-))
the initial question was whether or not there is an sRGB profile providing perceptual RI. Period. There is none; at least none that works as supposed to. Period.
No, not period...if you are going to do ANY post processing of any kind on one test image then you must do the same post on the other test image...
You CAN (and I often do) tweak an image in ProPhoto RGB while soft proofing in sRGB to massage and manipulate the conversion prior to actually converting. It's not the same as a more automated color transform using perceptual but it does offer control over the way the RelCol will impact the conversion.
-
No, not period...if you are going to do ANY post processing of any kind on one test image then you must do the same post on the other test image...
hups... I completely missed the part that there was further editing on the second image!
Sorry!!
Well, this is indeed apples to oranges...
-
hups... I completely missed the part that there was further editing on the second image!
Sorry!!
No problem...although I do happen to think the same post on the PP RGB image before going to sRGB in RelCol could have helped the results a bit. Which was why I though it was Apples to Oranges...
-
You CAN (and I often do) tweak an image in ProPhoto RGB while soft proofing in sRGB to massage and manipulate the conversion prior to actually converting. It's not the same as a more automated color transform using perceptual but it does offer control over the way the RelCol will impact the conversion.
No step was hidden,
it’s an open source (see attachment).
Peter
--
-
No step was hidden,
it’s an open source (see attachment).
But the point is that you did additional post processing to the second image that wasn't duplicated in any manner in the simple PP RGB to sRGB transform...one CAN (and should) soft proof a PP RGB image in sRGB prior to transform if you want to see what's gonna happen to color saturation and detail and do something about before doing the transform.
-
No problem...although I do happen to think the same post on the PP RGB image before going to sRGB in RelCol could have helped the results a bit. Which was why I though it was Apples to Oranges...
on the other hand he first converted perceptual to Photogamut and afterwards applied further adjustments. If there is already clipping in the converted file, there's not so much to improve with further editing...
Photogamut is very useful to map in particular bright high saturated colours into the range of sRGB (respectively in the range of printable colours)... so if you just want to "compress" the colours of an image without clipping it's an easy workflow (especially when you have to convert a number of different images...).