Luminous Landscape Forum
Raw & Post Processing, Printing => Colour Management => Topic started by: Doug Gray on February 25, 2016, 07:43:44 pm
-
A rather interesting difference between V2 and V4 profiles exists in Perceptual Intent. V4 profiles have a deep shadow cutoff at L=3.1. V2 profiles do not.
Relative Intent profiles are the same but have a curious sharp change in color around the paper's black point (L=8) where the highly neutral tone suddenly changes with a* going from 0 to just under +2. And, of course, shadows are blocked below L=8.
Perceptual intent doesn't maintain complete color neutrality but gradually shifts to the paper's black point.
-
Thanks Doug, but I've an ignorant question: are these ICC profiles created via a specific progrom+device, or are these features somehow hardwired into all B&W ICC's?
-
Well using Copra shows a tiny difference between V2 and V4 everything else being equal (these were CMYK profiles FWIW). I think Doug is using i1Profile which I also use. But I decided to try Copra to see if it too made a difference and it does by 1 device value in K and 2 in Yellow reported by Photoshop. I can pull out the old Macbook and run a test on MonacoPROFILER but I suspect we'll see the same thing. RelCol produces identical values with both V2 and V4 generated profiles.
V4 profiles are today, rather worthless anyway.
-
Thanks Doug, but I've an ignorant question: are these ICC profiles created via a specific progrom+device, or are these features somehow hardwired into all B&W ICC's?
These were ICC color profiles created with I1Profiler and Baryta neutral semigloss without OBs. The purpose was to show how black and white images would be rendered using standard profiles.
-
Well using Copra shows a tiny difference between V2 and V4 everything else being equal (these were CMYK profiles FWIW). I think Doug is using i1Profile which I also use. But I decided to try Copra to see if it too made a difference and it does by 1 device value in K and 2 in Yellow reported by Photoshop. I can pull out the old Macbook and run a test on MonacoPROFILER but I suspect we'll see the same thing. RelCol produces identical values with both V2 and V4 generated profiles.
V4 profiles are today, rather worthless anyway.
That's pretty much what I get. One of the curious aspects of the V4 profiles is the flat area below L=3 in graph 3 (V4 Perceptual). Drilling down into the profile the flat area is made by the "B Curve" of the BtoA0 tag. It cuts off exactly at L=3.1 which is the presumed black point of the ICC "reference medium." Per the spec, this only applies to Perceptual in V4 profiles in the BtoA0 tag. V4 profiles are rendered identically in the Relative tag (BtoA1). V2 tables have something similar named "input channels" and these are, in the ones I've looked at, always just a unity mapping.
Both V4 and V2 profiles implement a shallow "S curve" to enhance contrast in Perceptual which is appropriate for the assumed, standard reference medium.
Oddly, Photoshop either doesn't implement the B Curves in the BtoA0 tag or adds a bias to the PCSLAB values that obviates the BtoA0 clipping, probably the former. This can be seen by converting an L ramp from 0 to 10 using a V4 printer profile. Proper conversion using the V4 algorithms should incorporate the B Curve and clip the device outputs below L=3.1 for Perceptual Intent. Little CMS does implement this per the V4 spec. Many programs use LCMS so there can be some compatibility issues interchanging between these. Which gets back to your point about V4 profiles being iffy. V2 profiles suffer from a more ambiguous specification but it appears that profile vendors are now making fairly consistent V2 profiles and the extant software handles them more consistently. Also, we are a ways away from being able to use the new floating point and parametric conversions that V4 promises which should have better profiles with a smaller footprint. But when you dance with a gorilla you have to follow the big guy's steps or you get trod on.
EtoA: Turns out it is the latter (see above) . I modified a profile's B curve on the BtoA0 tag to test this. Adobe adds a bias on V4 profiles to the PCSLAB values which produces somewhat similar results as the V2 profiles and then processes the V4 profiles while LCMS does not add a bias. Interesting.
Perhaps the most important takeaway is that Relative Colorimetric (both V2 and V4) is rendered quite accurately within gamut (Nice straight line) but the way it is defined is that it scales from the White Point to L=0,0,0 and not the paper's black point. As a result, in the proximity of the print's Dmax, there is an algorithm shift where an attempt is made to map to the closest gamut point and that is the paper's black point, not the darkest neutral color. This results not only in deep shadow blocking but a sudden color shift as well. In the case of the Baryta paper this is an increase of just under 2 in the a* channel. I've seen other papers where the shift was as much as 4. While this is arguably the correct approach for Relative Colorimetric Intent it can produce unexpected results in deep shadows, particularly on matte papers with a low (poor) Dmax.
-
The algorithm Photoshop uses when converting to printer device in Perceptual with V4 profiles is this:
The image is first converted to XYZ. Then, the base black point at Lab(3.125,0,0) is converted to XYZ and the image XYZ values are scaled linearly between the base black point and the white point. This is then converted to Lab and run through the profile's BtoA0 LUTs. to produce RGB (or CMYK if required).
This introduces a small amount of error so when a conversion back to a working space is made the errors introduced are higher than those introduced when converting and inverting V2 profiles in Photoshop.
For instance, if I start with a 16 bit Lab(10,0,0) patch the and do a round-trip in Photoshop I get L=9.8 on a V2 profile and L=7.7 on a V4 profile for the same spectral data and settings with I1Profiler. The .2 difference in V2 profiles is within normal error margins from LUT interpolation. The difference of 2.3 in V4 profiles is not. It is due to inconsistencies in Adobe Photoshop's handling of V4 profiles.
-
Very interesting this treath.
But if all comments about V2 vs. V4 are right, why Eizo, X-Rite, and others manufacturers the default is ever V4?
-
But if all comments about V2 vs. V4 are right, why Eizo, X-Rite, and others manufacturers the default is ever V4?
Marketing I suspect (4 is larger than 2, must be better).
-
Or maybe follow the recomendations of International Color Consortium, as states on his web page: http://www.color.org/whyusev4.xalter
Why this controversy?, the passage of V2 to V4 seems an advance according to ICC.
And yes, I know that there are still software that does not properly implement the version 4.
Thanks!
-
Or maybe follow the recomendations of International Color Consortium, as states on his web page: http://www.color.org/whyusev4.xalter (http://www.color.org/whyusev4.xalter)
Why this controversy?, the passage of V2 to V4 seems an advance according to ICC.
And yes, I know that there are still software that does not properly implement the version 4.
NO v4 profile I know of follows or uses the PRMG. Therefore, they are V4 profiles in sheep's clothing. What advances are you finding?
-
NO v4 profile I know of follows or uses the PRMG. Therefore, they are V4 profiles in sheep's clothing. What advances are you finding?
Indeed, this is rather interesting. There is a tag when Perceptual uses PRMG and the I1Profiler V4 profiles do not use that tag. It's optional in V4. But here's the rub. One of the benefits claimed for V4 is better invertability but, as shown above, Photoshop produces materially more error inverting a Perceptual transform than does the V2 profile. The I1P v4 profiles do implement a clip below L=3.1 in Perceptual which is exactly the black point of the PRMG even though they don't use the PRMG tag. Perhaps this is why Photoshop inverts V4s less accurately. Who knows?
That said, by far the major benefits V4 produced are better colorimetric specifications which V2 makers were urged to, and have been, implementing. In particular some of the older V2 profiles altered the way Colorimetric conversions were done. The most egregious, from my POV, were the scaling of black points in the BtoA1 tables, a sort of black point correction. This made RelCol more similar in typical use to Perceptual but at the cost of losing colorimetric accuracy. The tradeoff was poorer soft proofing. With the V4 spec this was more clearly defined as strictly colorimetric so similar results can be obtained using Adobe's BPC in RelCol without screwing up soft proofing or cross rendering.
Unfortunately, the major improvements that V4 could produce are from the new parametric and floating point tables. These would allow more accurate results, especially near gamut edges and with highly nonlinear devices. The current LUT structures intrinsically produce errors in the gamut transition regions. I1Profiler does not implement most of these new structures in V4.
At this point I consider V2 profiles to have more predictable and consistent properties but with the caveat that the V2 spec is more, um, "flexible," so one has to be careful using older V2 profiles. I've only seen this with some canned printer profiles. I don't use them so it isn't a problem.
-
Hi digitaldog,
International Color Consortium says on his web:
================
Benefits that users will see with v4 are:
•More consistent and accurate colorimetric intent transforms resulting from a fully-defined media white point and more consistent handling of black points
•More consistent and higher quality perceptual intent transforms because there is now a fully-defined perceptual reference medium dynamic range and gamut
•Smaller and more accurate profiles that make use of new efficient transform structures
•Generally improved consistency between different profiles and CMMs as a result of improvements to the specification throughout
•Better support for re-purposing and proofing of images rendered with the perceptual and colorimetric intents through more accurate inverse transforms
===============
I do not have enough experience in Color Management (just I read your book!) to contradict you or ICC, but I can say that for years I use in all of my workflow V4, in my monitors, profilers and printers, achieving very good results in soft-proofing.
I am willing to change to version 2 if I have good arguments.
Thanks!
-
Hi digitaldog,
International Color Consortium says on his web:
================
Benefits that users will see with v4 are:
•More consistent and accurate colorimetric intent transforms resulting from a fully-defined media white point and more consistent handling of black points
•More consistent and higher quality perceptual intent transforms because there is now a fully-defined perceptual reference medium dynamic range and gamut
•Smaller and more accurate profiles that make use of new efficient transform structures
•Generally improved consistency between different profiles and CMMs as a result of improvements to the specification throughout
•Better support for re-purposing and proofing of images rendered with the perceptual and colorimetric intents through more accurate inverse transforms
===============
I do not have enough experience in Color Management (just I read your book!) to contradict you or ICC, but I can say that for years I use in all of my workflow V4, in my monitors, profilers and printers, achieving very good results in soft-proofing.
I am willing to change to version 2 if I have good arguments.
Thanks!
What they say a properly and fully implemented V4 profile should provide, and what Doug and I have illustrated with two different packages don't sync up. Simple as that!
The ICC as a body itself doesn't build any profile packages I'm aware of. Some of their members do. And yes, you probably should consider a revert to V2 because V4 provides nothing useful that I'm aware of and can cause issues with some software products.
-
What they say a properly and fully implemented V4 profile should provide, and what Doug and I have illustrated with two different packages don't sync up. Simple as that!
The ICC as a body itself doesn't build any profile packages I'm aware of. Some of their members do. And yes, you probably should consider a revert to V2 because V4 provides nothing useful that I'm aware of and can cause issues with some software products.
I agree with Rodney.
I would love to see some of the promised benefits of V4 profiles such as parametric curves that would improve accuracy without requiring large, LUT laden, profiles. It hasn't happened. So long as V4 profiles are really mostly V2 profiles in sheep's clothing, as Rodney puts it quite accurately, why use profiles that may be interpreted oddly by some software.
Keep in mind that the V4 profiles were introduced well over a decade ago and the promised improvements, while nice, haven't yet materialized - with the exception that profile makers have adopted the more specific measurement based criteria and Bradford Adaption of V4. Since these are fully compatible, even urged, for V2 profiles why not just use V2 and avoid any potential compatibility issues.
Maybe in time the new V4 features will come of age and be widely consistently incorporated but it's sure been slow so far.
To be fair, the differences we see are minor, but when Photoshop V4 profiles convert then invert a B&W neutral tone curve less accurately than a V2 profile it is not a step forward.
-
Why this controversy?, the passage of V2 to V4 seems an advance according to ICC.
It's natural that after all their hard work in creating it, that they are advocating it.
That doesn't mean that there are persuasive technical reasons for adopting V4 over V2 in most ordinary circumstances though.
And yes, I know that there are still software that does not properly implement the version 4.
V4 added a level of complexity, particular in regard to the change in L*a*b* PCS encoding, and black point handling.
-
NO v4 profile I know of follows or uses the PRMG. Therefore, they are V4 profiles in sheep's clothing. What advances are you finding?
Use and marking the use of the PRMG is optional. And there's actually no reason that V2 profiles can't make use of the PRMG if they want to either.
-
Use and marking the use of the PRMG is optional. And there's actually no reason that V2 profiles can't make use of the PRMG if they want to either.
Well, as the ICC and some of it's members describe it, the PRMG sounds like it could be rather useful. Yet no matter the version, who's using it and why not?
-
One of the benefits claimed for V4 is better invertability
Really ? Where ?
The I1P v4 profiles do implement a clip below L=3.1 in Perceptual which is exactly the black point of the PRMG even though they don't use the PRMG tag.
L=3.1 is the black point of the PRM, independent of the use or not of the PRMG.
That said, by far the major benefits V4 produced are better colorimetric specifications which V2 makers were urged to, and have been, implementing. In particular some of the older V2 profiles altered the way Colorimetric conversions were done. The most egregious, from my POV, were the scaling of black points in the BtoA1 tables, a sort of black point correction. This made RelCol more similar in typical use to Perceptual but at the cost of losing colorimetric accuracy. The tradeoff was poorer soft proofing. With the V4 spec this was more clearly defined as strictly colorimetric so similar results can be obtained using Adobe's BPC in RelCol without screwing up soft proofing or cross rendering.
The scaling of the colorimetric table black point was not something that was specified or even allowed in the V2 specification (i.e. See ICCV2.4 page 66 Equations A.1 to A.3), irrespective of how some may have mis-implemented it, so this is not something that V4 specification has really brought to the table. There are certainly many V2 profile makers that don't have this behavior.
Unfortunately, the major improvements that V4 could produce are from the new parametric and floating point tables. These would allow more accurate results, especially near gamut edges and with highly nonlinear devices. The current LUT structures intrinsically produce errors in the gamut transition regions. I1Profiler does not implement most of these new structures in V4.
I don't see this to be the case at all for printer profiles. As I understand it, the parametric and floating point tables are of primary value for synthetic and display & camera profiles, where the device behavior can be more mathematically described, and extreme dynamic range is involved. Printers are inherently less well behaved, hence requiring an arbitrary mapping described by a lookup table. Given the dynamic range and precision of printer sample readings, floating point encoding provides little benefit over 16 bit encoding, and some disadvantage in terms of profile size.
At this point I consider V2 profiles to have more predictable and consistent properties but with the caveat that the V2 spec is more, um, "flexible," so one has to be careful using older V2 profiles. I've only seen this with some canned printer profiles. I don't use them so it isn't a problem.
I don't disagree with this assessment.
-
I would love to see some of the promised benefits of V4 profiles such as parametric curves that would improve accuracy without requiring large, LUT laden, profiles. It hasn't happened.
Nor are you ever likely to, for printer profiles.
Synthetic display profiles like sRGB etc, yes, because the parametrized curves let the straight segment of the sRGB luminance curve be more efficiently encoded than a LUT, which was the primary reason for adding this tag.
-
Well, as the ICC and some of it's members describe it, the PRMG sounds like it could be rather useful. Yet no matter the version, who's using it and why not?
I've not tried to keep up with it, but I know at one stage HP made some profiles using the PRMG.
The main technical problem with the PRMG is the same one that's always been with the idea of pre-computed A2B and B2A tables:
you can't implement a full range of intents properly. Anything using the PRMG (or any other common intermediate gamut where the source and destination profiles are arbitrarily mixed) is going to end up with what I would call a saturation intent. That's because all you can really do is map the source space gamut to the PRMG, and then the PRMG to the destination spaces gamut. So if the destination has a larger gamut in some places than the source, you get gamut expansion. In practice this may not be much of an issue for typical RGB display space to printer space, but it makes it a not-so-general gamut mapping strategy.
Another issue I've noted before, is that you get two gamut mapping transforms being concatenated. Each transform involves a degree of inaccuracy, so the inaccuracy is doubled. And if the two profiles are created by different software, there is no guarantee that the two transforms cancel out very well where they should.
-
Nor are you ever likely to, for printer profiles.
Synthetic display profiles like sRGB etc, yes, because the parametrized curves let the straight segment of the sRGB luminance curve be more efficiently encoded than a LUT, which was the primary reason for adding this tag.
Parametric curves do reduce the size of sRGB profiles since 1D linear LUTs aren't needed. Data rates were slower, storage was more expensive, and images smaller back in 2001 when V4 was first introduced. The few KBs difference is far less material these days.
However, the floating point 3DLUTs in V4.3 are much improved. They are no longer forced into having the same number of grid points in each dimension and have more preprocessing capability with matrix transforms on both ends to get the most out of the LUTs. That can make a big difference for the same level of accuracy. OTOH, most printer profiles aren't attached to images and the V2 16 bit 3D LUTs can be made as accurate as desired. Size is hardly an issue these days. These changes primarily benefit print devices and were put in to improve them.
-
I've not tried to keep up with it, but I know at one stage HP made some profiles using the PRMG.
The main technical problem with the PRMG is the same one that's always been with the idea of pre-computed A2B and B2A tables:
you can't implement a full range of intents properly. Anything using the PRMG (or any other common intermediate gamut where the source and destination profiles are arbitrarily mixed) is going to end up with what I would call a saturation intent. That's because all you can really do is map the source space gamut to the PRMG, and then the PRMG to the destination spaces gamut. So if the destination has a larger gamut in some places than the source, you get gamut expansion. In practice this may not be much of an issue for typical RGB display space to printer space, but it makes it a not-so-general gamut mapping strategy.
Another issue I've noted before, is that you get two gamut mapping transforms being concatenated. Each transform involves a degree of inaccuracy, so the inaccuracy is doubled. And if the two profiles are created by different software, there is no guarantee that the two transforms cancel out very well where they should.
It's never likely to be resolved. It isn't even clear to me how it could be done. Interesting idea though. In the abstract.
-
Really ? Where ?L=3.1 is the black point of the PRM, independent of the use or not of the PRMG.
You are right. L=3.1 is the PRM either way. I worded it incorrectly. The problem is that Adobe Photoshop, doing a Perceptual->Device->Perceptual, using only a neutral tone curve, produces significantly more inversion error using V4 profiles than V2 profiles. Adobe adds in an L=3.1 base level before going to the profile which then clips it out. So the printer is printing close to the same way V2 profiles print. However, when doing an inversion, consideration of the PRM is not done resulting in increased error. See the attached plots of the inversion error in V2 and V4 Perceptual. I believe the V4 profiles are created correctly and the increased error is from Photoshop.
The scaling of the colorimetric table black point was not something that was specified or even allowed in the V2 specification (i.e. See ICCV2.4 page 66 Equations A.1 to A.3), irrespective of how some may have mis-implemented it, so this is not something that V4 specification has really brought to the table.
The ICC says there were ambiguities in spec'ing the V2 Colorimetric tables and V4 specifically clarifies that:
http://www.color.org/advantagesv4.pdf
In v2 profile builders were allowed to modify measurement data prior to building the relative colorimetric tables for a profile. This sometimes led to differences in the way in which colorimetric data could be interpreted when a colorimetric match is required.
I credit the V4 development with correcting this ambiguity and further, strongly urging V2 profile makers to do the same. For the most part it seems they have. As I said, I've only seen this screwed up BtoA1 table in older canned profiles.
Neutral tone curve error of Perceptual Roundtrip of L=0 to L=100 for V2 and V4 profiles using Photoshop CC x64 in Win 10. Orange is the V4 error, blue is the V2 error. I1Profiler.
L values were in increments of .1. The X axis data is labeled 10x actual L values. Too lazy to scale the axis.
-
However, the floating point 3DLUTs in V4.3 are much improved. They are no longer forced into having the same number of grid points in each dimension and have more preprocessing capability with matrix transforms on both ends to get the most out of the LUTs. That can make a big difference for the same level of accuracy. OTOH, most printer profiles aren't attached to images and the V2 16 bit 3D LUTs can be made as accurate as desired. Size is hardly an issue these days. These changes primarily benefit print devices and were put in to improve them.
Marginal improvements at best, so return on investment in using them (by creating software that emits them) is low.
-
Well, if Digitaldog, Doug Gray and GWGill, believe that today the version 2 is the best option, what I can add?
But is hard to understand that a company like X-Rite, can advise on the i1Profiler's Help that, except in case of incompatibility, the best option is version 4.
X-Rite says textually "In general, version 4 profiles will give greater precision"
Marketing perhaps, according to Digitaldog?
Reviewed the comments, I'm thinking of changing my profiles to version 2.
Thanks!
-
Marginal improvements at best, so return on investment in using them (by creating software that emits them) is low.
True. As it is accuracy is quite good at the inner parts of a gamut with 16 bit LUTs. The iffy areas are near the gamut edges and that can be reduced easily if needed by expanding the grid point size. Storage is cheap and printer profiles aren't usually attached to images. Even where they are images have grown so much in the last decade it makes the profile size small by comparison. I've also noticed that the vendors take the shortest path to being able to label profiles "V4." I've never seen a shaping matrix, newly part of V4 3D LUT paths, that is anything other than pass through. So they just take the same V2 data and plop it into V4 with the PRM front end change for Perceptual and voila, they have a profile they can call V4.
I mostly use Rel Col and that is identical in I1Profiler V2 and V4 profiles so why switch.
-
Well, if Digitaldog, Doug Gray and GWGill, believe that today the version 2 is the best option, what I can add?
What I recommend you do is build a few profiles both ways, run some tests.
-
Well, if Digitaldog, Doug Gray and GWGill, believe that today the version 2 is the best option, what I can add?
Since V4 is essentially a super-set of V2, there's no reason they can't work just as well, but getting inter-interoperability right is hard, particularly when a vendor gets locked into something by not wanting to cause incompatibility with it's own previous versions.
-
Hi all,
I tested versions 2 and 4 of some profiles and the truth is that the differences are almost negligible.
Higher out of gamut of the V4 (very small), slightly higher volume for V2 is slightly seen and slightly diferences in Dmax for V4 (max 0,02).
Maybe in the Round Trip Colorimetric is where a difference can be seen for V2. Attached is an example Hahnemuhle Fine Art Baryta.
-
Hi all,
I tested versions 2 and 4 of some profiles and the truth is that the differences are almost negligible.
Higher out of gamut of the V4 (very small), slightly higher volume for V2 is slightly seen and slightly diferences in Dmax for V4 (max 0,02).
Maybe in the Round Trip Colorimetric is where a difference can be seen for V2. Attached is an example Hahnemuhle Fine Art Baryta.
In Perceptual (BtoA0) only the L curve is clipped in V4 (starts ramping at L=3.1) but the a* and b* color portions are not so this could produce small differences as well in the non-neutral areas. It might also affect "gamut volume," a rather nebulous concept outside of Absolute Colorimetry. I've only investigated Perceptual and Colorimetric intents for B&W purposes. I don't see any difference at all in Colorimetric roundtrips. The differences in Perceptual are minimal. What I don't see is any reason to use V4 printer profiles at this time but I doubt V4 profiles would be visually different. At least on the I1Profiler.