ICC Profile Choice - ProPhoto RGB

[AlterEgo]

...

for as long as we do not use input and output profiles that clamp the data during color transforms (and both profiles are used sequentially at the end of the pipeline in C1) we can... it seems nothing internally in C1 by itself does such clamping before a final stage where a color transform guided by some input profile (camera profile) is applied, then followed by a color transform guided by some output profile... with a fine print also that we can also actually alter the color transfer guided by an input profile through C1 color editor for as long as that input profile conforms to P1 rules - LAB PCS and A2B0 table...

[digitaldog]

again wrong question. To edit in ProPhoto you don't have to import the profile into C1. As long as it is installed on the system so that you can set it as output space in C1 you are effectively editing in ProPhoto-RGB!

I can't speak for the OP, that bit of course makes sense to me. I don't know that's what he's referring to. As I read it, he's saying he's bringing in a rendered image (?) in ProPhoto RGB (Can I import ProPhoto RGB and make it an option to edit in?) and then, use that working space for editing. But that's just a guess. I then asked, does the processing color space match somewhat or exceed ProPhoto RGB as it would be pointless to ask or need it if the color space were significancy smaller. That got us into this long but useful discussion. If I'm to believe what I've read (much of it from those outside PhaseOne), the answer is yes, whatever processing color space gamut is going on under the hood, it isn't a 'waste' (as it would be in say Aperture) to be wanting to edit existing ProPhoto RGB data.

[tho_mas]

I can't speak for the OP, that bit of course makes sense to me. I don't know that's what he's referring to. As I read it, he's saying he's bringing in a rendered image (?) in ProPhoto RGB (Can I import ProPhoto RGB and make it an option to edit in?) and then, use that working space for editing.

My reading of the initial question was different since he started with asking how to set the "Base Characteristics" (which is where you can [but don't have to!] choose the input profile). So to me it was clear he was asking how to set ProPhoto-RGB as input profile... but of course I might be wrong ...

However, you bring up a very weak point of C1. Editing of TIFs and JPEGs. When you load ("import") e.g. a TIF file into C1 the embedded profile of the image gets ignored. Instead C1's internal generic "neutral" color space gets assigned. While visually identical in so far the image does not contain colors that exceed C1's "neutral" color space all colors outside of C1's neutral color space get desaturated. This is - at best - a half backed solution and it's clearly better NOT to use C1 for editing TIFs/JPEGs!

[ErikKaffehr]

Hi,

I have rechecked and it is clear that ProPhoto RGB does indeed not cover all visible colours, but it nicely encloses a colour space called Pointer's Gamut. "The Pointer’s gamut is (an approximation of) the gamut of real surface colors as can be seen by the human eye, based on the research by Michael R. Pointer (1980). What this means is that every color that can be reflected by the surface of an object of any material is inside the Pointer’s gamut."

http://www.tftcentral.co.uk/articles/pointers_gamut.htm

To make a fine point, what is the benefit of the advanced colour editor pushing colours outside the colour gamut of any monitor or any printing device. Those are colours that cannot be represented by any output device? Also the colours cannot be observed as the display will clip them.

I have actually observed this in one of my experiments with C1, it pushes deep blues that are perfectly inside Adobe RGB wide outside Adobe RGB, causing a large colour reproduction error although it is not visible on a screen with Adobe RGB as it is clipped to Adobe RGB in visualisation.

This example was taken from a P45+ image of an IT-8 test chart with all samples inside Adobe RGB and processed with Capture One v7 without any colour editing.

Best regards
Erik

No nonsense. Read correctly: you can produce colors that exceed ProPhoto. To do this you can increase global saturation in the Advanced Color Editor. But by default - so without any color edits - the input profiles are typically somwhere between Adobe-RGB and ProPhoto-RGB ... attached the default input profile of your camera compared to ProPhoto-RGB...

[tho_mas]

To make a fine point, what is the benefit of the advanced colour editor pushing colours outside the colour gamut of any monitor or any printing device. Those are colours that cannot be represented by any output device? Also the colours cannot be observed as the display will clip them.

Oh my god. Erik - I could boost global saturation of a real world image by 200% - but why should I do so?
It seems, the "Advanced Color Editor" is aimed at advanced users...
It is certainly not desinged to produce imaginary colors (altough it could/can - just as Adobe software with the huge color space ProPhoto).
Think of a certain very, very desaturated color that you want to push... maybe even by 200% ... got it?
The capabilities of a certain tool do not dictate how to use this very tool ... it's always the end-user that has to make the most out of it.
Again: the gamut of C1's input profiles is typically somewhere between Adobe-RGB and ProPhoto-RGB ... and normally there's no need to push the Advanced Color Editor to its limits. Normally you do very subtle edits (at least I do).

[Dan Wells]

Could the internal color space be so large as to fully contain all supported input and output color spaces? Realistically, a color space only a little larger than ProPhoto should do this - very little of value is outside ProPhoto, although Joseph Holmes' DCam 5 is even larger, and is supposed to fully encompass human vision. No monitor I know of can display any color outside of ProPhoto, and no printer can print one... Some monitors CAN display a few colors outside of Adobe RGB, and 11-12 ink printers can print substantially outside of Adobe RGB, but neither can hit the edges of ProPhoto. Can some weird, reflective color we see on occasion (a brand-new traffic cone?) get outside of ProPhoto? I don't know. Can a flower display colors outside of ProPhoto to a bee? Almost for sure - some flowers reflect UV, and bees can see it. That could even be outside of DCam 5, but since we can't see it or print it, why worry about it unless you are hanging a show in an apiary...

If C1 worked internally in something like DCam 5, why would it clip when converting to (or from) anything? As long as it's always 16 bits/channel, quantization errors are unlikely (and might it be 32 bits/channel, eliminating them entirely)? That would more or less allow it to claim to be space-neutral internally - it imports images in in the camera space, displays them in monitor space, exports in a selected export space and prints in printer space, all of which fit within the huge internal space. Lightroom almost does this with MelissaRGB (it's very hard, but not impossible, to find a color outside of that space). What if C1's space were even larger? Could that make it truly neutral internally (although highly dependent on rendering intent as it converted)?

[AlterEgo]

Could the internal color space be so large as to fully contain all supported input and output color spaces? Realistically, a color space only a little larger than ProPhoto should do this - very little of value is outside ProPhoto, although Joseph Holmes' DCam 5 is even larger, and is supposed to fully encompass human vision. No monitor I know of can display any color outside of ProPhoto, and no printer can print one... Some monitors CAN display a few colors outside of Adobe RGB, and 11-12 ink printers can print substantially outside of Adobe RGB, but neither can hit the edges of ProPhoto. Can some weird, reflective color we see on occasion (a brand-new traffic cone?) get outside of ProPhoto? I don't know. Can a flower display colors outside of ProPhoto to a bee? Almost for sure - some flowers reflect UV, and bees can see it. That could even be outside of DCam 5, but since we can't see it or print it, why worry about it unless you are hanging a show in an apiary...

If C1 worked internally in something like DCam 5, why would it clip when converting to (or from) anything? As long as it's always 16 bits/channel, quantization errors are unlikely (and might it be 32 bits/channel, eliminating them entirely)? That would more or less allow it to claim to be space-neutral internally - it imports images in in the camera space, displays them in monitor space, exports in a selected export space and prints in printer space, all of which fit within the huge internal space. Lightroom almost does this with MelissaRGB (it's very hard, but not impossible, to find a color outside of that space). What if C1's space were even larger? Could that make it truly neutral internally (although highly dependent on rendering intent as it converted)?

did you actually read the topic, specifically the quote from P1 developer ?

[Dan Wells]

At least as I'm reading it, it's entirely consistent - they keep the camera space, but have it WITHIN something huge, so that "the gamut you can see in the camera space does not limit the gamut used for internal processing". If this weren't true, you could never do highly saturated HDR (no camera can record some of those colors). Since C1 can convert to ProPhoto for output, and, from the developer's post, it sounds like it can do it without clamping, the internal processing gamut (maybe it's a gamut, not a true space) has to contain ProPhoto...

[ErikKaffehr]

Hi,

I wouldn't say god is involved in this…

In the experiments I made I found that Capture One pushes deep blues well inside Adobe RGB into well outside Adobe RGB blues. Now, most computer screens used sRGB which are based on REC 709 RGB. High end wide gamut devices are close to Adobe RGB. But neither colour space includes all colours occuring in nature, which is often referred to as  Pointer's RGB.

So if colours are pushed outside Adobe RGB, colour differences will not be visible on a computer monitor.

Check the enclosed figures at the bottom of this page:

The first figure shows how colours are shifted when an IT-8 produced by chromogenic printing is  reproduced by a P45+ back and interpreted in Capture One. As you can see the colours have a significant shift.

The second figure shows that the colours are pushed well outside Adobe RGB.

The third figure shows that although colour error is pretty large, about Delta E = 17, the difference is barely noticeable on screen, for the simple reason that Adobe RGB cannot correctly show the real colour.

The forth figure demonstrates that a much smaller colour error is very much visible if both the reference and sampled colour lie within the gamut visible within Adobe RGB.

Now, there is a new colour space defined for 4K called Rec 2020. It is possible that the colours produced by Capture One default processing would be visible in Rec 2020. I don't know, because I have no Rec 2020 screen.

---------------------------------

Another issue which may be related or not was found when I made a test looking at reproduction of greens. To a great part because I had some communication with Tim Parkin, publisher of OnLandscape. Tim indicated that he and his friend Joe Cornish had issues with colour reproduction on the P45+, with greens being pushed towards yellows. I had no communication with Joe Cornish directly but Tim said that they both found that P45+ images needed massive corrections for realistic colour. Tim and Joe also found that the later generation DALSA based backs delivered better colour.

What I wanted to find out how much that issue was related to profiles and how much related to sensor. One of the experiments I set up was a blush purple flower with green blades. I have chosen this because I knew these colours were sensitive to reproduction errors. Illumination was with Elinchrome studio flash.

Capture One gave these colours:


While other alternatives essentially yielded these colours:


Now, I measure the colours on the flower with a spectrometer and found the colours were these:


With the green blades it was a bit different, a lot more variation between samples.


So, Capture One essentially reproduced bluish purple flowers as blue. This may or may not correspond to the blue colour shift mentioned above.

I did not really come up with a clear conclusion. One thing I observed the IR content on both green and the purple flowers was very high. So the amount of IR filtering may affect colour reproduction and that may be hard to correct in processing.

Finally, if we look at colour reproduction from a Sony Alpha 99 the colour from C1 is still very blue:


While Lightroom reproduction is close to original:


Now, I am perfectly aware that the colour can be easily adjusted in the Capture One colour editor, but I feel that a good raw converter should deliver decent colours by defaults.

A more comprehensive posting is here: http://echophoto.dnsalias.net/ekr/Articles/OLS_OnColor/SimpleCase/

A draft of the article I was writing for On Landscape is here: http://echophoto.dnsalias.net/ekr/Articles/OLS_OnColor/OnColor.pdf

The article never got published, I guess that both I and Tim run out of steam and now clear conclusion could be drawn. An interesting observation may be that Tim Parkin is also a painter. Painters work on location and try to reproduce what they see using available paints. I think that kind of viewing produces an awareness of colour that we photographers shooting in landscape and processing in the digital darkroom will not see.

Anders Torger, a frequent poster here, has developed a new tool called DCamprof which can generate both DNG profiles for Lightroom and ICC profiles for Capture One. It is described here: http://www.ludd.ltu.se/~torger/dcamprof.html . Those pages offer a lot of insight in both profiles and interpretation of colour.

Best regards
Erik

Oh my god. Erik - I could boost global saturation of a real world image by 200% - but why should I do so?
It seems, the "Advanced Color Editor" is aimed at advanced users...
It is certainly not desinged to produce imaginary colors (altough it could/can - just as Adobe software with the huge color space ProPhoto).
Think of a certain very, very desaturated color that you want to push... maybe even by 200% ... got it?
The capabilities of a certain tool do not dictate how to use this very tool ... it's always the end-user that has to make the most out of it.
Again: the gamut of C1's input profiles is typically somewhere between Adobe-RGB and ProPhoto-RGB ... and normally there's no need to push the Advanced Color Editor to its limits. Normally you do very subtle edits (at least I do).

[bjanes]

Yes, they're called imaginary colors, but you know that. According to Bruce Lindbloom, ProPhoto RGB has a coding efficiency of 87.3% which means that it allows to encode 12.7% of imaginary colors on top of the 'real colors' (humanly visible, on average).

Those imaginary colors are mathematically usable and may still be relevant when one e.g. locally shrinks the saturation so that these colors become 'real colors'.

Cheers,
Bart

Lindbloom also lists values for Lab gamut efficiency, which is the percentage of the Lab gamut that a given space can encode. The value for ProPhotoRGB is 91.2%, which means that ProPhotoRGB can not encode 8.8% of visible colors. This is partially illustrated by the shown figure, which is taken from the paper regarding Pointer's work that Erik Kaffehr has cited.

That said, ProPhotoRGB does encode all of the real world surface colors (shown by the irregular space shown within the ProPhoto triangle) by a comfortable margin, whereas AdobeRGB does not. This is why many prefer ProPhotoRGB as one's working space.

Bill

[ErikKaffehr]

Hi Bill,

Thanks for your lucid explanation!

It would be interesting to see a gamut plot of the Rec 2020 RGB compared to the ProPhoto RGB, as Rec 2020 is around the corner. But I have not found a Rec 2020 profile to plot.

My understanding is that imaginary primaries are needed to enclose a large part of the visible colours. The full range of colours can not reproduced with three primaries with spectral colours. It would take quite a few spectral colour primaries to enclose the visible colours. The pointer's gamut is enclosed in Prophoto RGB and just somewhat clipped by Rec 2020, which uses three real world primaries.

Best regard
Erik

Lindbloom also lists values for Lab gamut efficiency, which is the percentage of the Lab gamut that a given space can encode. The value for ProPhotoRGB is 91.2%, which means that ProPhotoRGB can not encode 8.8% of visible colors. This is partially illustrated by the shown figure, which is taken from the paper regarding Pointer's work that Erik Kaffehr has cited.

That said, ProPhotoRGB does encode all of the real world surface colors (shown by the irregular space shown within the ProPhoto triangle) by a comfortable margin, whereas AdobeRGB does not. This is why many prefer ProPhotoRGB as one's working space.

Bill

[AlterEgo]

But I have not found a Rec 2020 profile to plot.

\Argyll\ref\Rec2020.icm

[bjanes]

\Argyll\ref\Rec2020.icm

Sorry, but that does not appear to be a valid URL. At least, it doesn't produce results with my system.

Bill

[AlterEgo]

Sorry, but that does not appear to be a valid URL. At least, it doesn't produce results with my system.

Bill

it is not an url , it is a location of that profile when you install argyll suite.

[Dan Wells]

Interesting that the problem colors seem to be violet-colored flowers, and many flowers have high UV reflectivity. How much can cameras "see" into the ultraviolet? Could C1 and Lightroom be responding to ultraviolet or near-UV differently?

[ErikKaffehr]

Thanks!

Erik

it is not an url , it is a location of that profile when you install argyll suite.

[digitaldog]

It would be interesting to see a gamut plot of the Rec 2020 RGB compared to the ProPhoto RGB, as Rec 2020 is around the corner. But I have not found a Rec 2020 profile to plot.

If you have the spec's (chromaticity values for RGB, white point, gamma), you can build it in Photoshop's Color Settings.

[Bart_van_der_Wolf]

If you have the spec's (chromaticity values for RGB, white point, gamma), you can build it in Photoshop's Color Settings.

That's right for the Rec. 2020 colorimetry. There is also a tone curve adjustment needed to get even closer to the recommendation. Plenty of bits per channel are also seriously recommended.

Cheers
Bart

[digitaldog]

That's right for the Rec. 2020 colorimetry. There is also a tone curve adjustment needed to get even closer to the recommendation.

True, PS can't create a TRC, just a simplified gamma value, but if the idea is to examine the gamut and one doesn't have access to the actual profile (for whatever reason), it's a decent start. I just did this today in Photoshop to get a better idea of how Apple's new displays which are based on DCI-P3 compare to Adobe RGB (1998).

[ppmax2]

Which is rendered in wireframe and which is rendered in shaded mode?

Thx
PP