What are the essential adjustments that SHOULD be done in the raw processor?

[tho_mas]

So for me it's still going to be sRGB first,  Adobe RGB second, Beta RGB third ... depending on the image and the destination ... and then if I absolutely need it, ProPhoto RGB (or perhaps the printer profile if I know precisely what the printer and paper are going to be).

on the one hand you are asking what's essential in RAW-processing ... on the other hand you won't use the full potential of a RAW processor by choosing totally irrelevant and dated icc Profiles that also may be too small to contain the all information contained in the RAW file.
sRGB refers to the gamut of TV-Monitors. However, it has its justification even today as you can regard it as the smallest denominator of color management... and images converted to sRGB do look "reasonable" on most computers even in applications without color management since many computer monitors have a gamut similar to sRGB (yet). AdobeRGB once was intended as a prepress color space but the color space actually has nothing to do with the gamut of real printers (it's a pure academic color space) and above all the Gamma 2.2 is the worst choice of an TRC with regard to a prepress workflow (at least in 8bit). Now, and BetaRGB was a very good concept for a working space encompassing the colors of different film types and all printable colors but I suspect the profile will never be updated to ICC-V4 specs. Too, its gamut is very similar to Rec2020 and therefore I would use Rec2020 as it's more common.

There are soooo many applications images may be used in today that it doesn't make sense to process an image with regard to a certain target in the first place. It makes much more sense to process a photo so that it contains all relevant data (a processed 1:1 copy of the adjusted RAW-file, if you want so...). And then convert to a certain color space with the help of gamut warning and the Info-Platte in Photoshop (or elswhere).

Just for reference... here's a recommendation of the ICC: http://www.color.org/prmg_gamutwarning.xalter ->

The ICC recommends the use of the ISO22028-2_ROMM-RGB.icc profile as the PRM working space and exchange encoding, and the PRMG_RGB-sRGB_based.icc profile as a gamut warning profile to check for the use of ROMM RGB colors outside the PRMG.

If you think about it I would say a reasonable advise is to use a color space for "original photos" that does at least encompass the Perceptual Reference Medium Gamut. You don't have to use ProPhoto-RGB (said ISO22028-2_ROMM-RGB profile is a ICC-V4 version of ProPhoto-RGB)... but AdobeRGB will certainly not do... If I would work with Lightroom (or ACR) I would defenitely process to ProPhoto-RGB and would convert the "original file" to any other color space in Photoshop (again: with the help of the gamut warning tool and the Info-Platte).

[digitaldog]

So I think that it's a good discipline to ask oneself, before rendering an image to a color space, or converting to another color space, what the destination of this particular variant of the image is and whether or not the image gamut will fit into the intended space. 

In an Adobe raw workflow, doesn't matter! Use ProPhoto RGB:

sRGB urban legend Part 1


In this 30 minute video I'll cover:
Is there benefit or harm using a wider color gamut working space than the image data?
Should sRGB image data always be encoded into sRGB?
What are RGB working spaces and how do they differ?
What is Delta-E and how we use it to evaluate color differences.
Color Accuracy: what it really means, how we measure it!
Using Photoshop to numerically and visually see color differences when using differing working spaces.
Using ColorThink Pro and BableColor CT&A to show the effects of differing working space on our data and analyzing if using a smaller gamut working space is necessary.
Appendix: testing methodology, how differing raw converters encode into working spaces, capturing in-camera JPEG data and color accuracy.


Low resolution (YouTube): https://www.youtube.com/watch?v=1w0zUIl-dzY
High resoution: http://digitaldog.net/files/sRGBMyths.mp4

[bjanes]

In an Adobe raw workflow, doesn't matter! Use ProPhoto RGB:

sRGB urban legend Part 1


In this 30 minute video I'll cover:
Is there benefit or harm using a wider color gamut working space than the image data?
Should sRGB image data always be encoded into sRGB?
What are RGB working spaces and how do they differ?
What is Delta-E and how we use it to evaluate color differences.
Color Accuracy: what it really means, how we measure it!
Using Photoshop to numerically and visually see color differences when using differing working spaces.
Using ColorThink Pro and BableColor CT&A to show the effects of differing working space on our data and analyzing if using a smaller gamut working space is necessary.
Appendix: testing methodology, how differing raw converters encode into working spaces, capturing in-camera JPEG data and color accuracy.

Andrew,

I looked at your video and found it well done and informative. Thanks for producing it. Without reviewing the video, I am not sure if your tests were done in 8 bit or 16 bit (15+1 in Photoshop). One would expect greater rounding errors in 8 bit. You do advise rendering into 16 bit ProPhoto, but what happens if you render into 8 bit? Some gurus say the differences between the larger and smaller color spaces could be exaggerated with drastic edits, even if clipping is avoided. Of course, a drastic edit would produce ΔEs from the reference image, so an accuracy criterion would no longer be applicable even if reference values were available.

Some gurus advocate editing in 32 bit floating point in specialized software. As far as I know PS, ACR, and LR edit internally in 16 bit, and does this make any practical difference.

These topics are interesting, but for practical work I find using 16 bit ProPhotoRGB is very satisfactory and this is what I use.

Regards,

Bill

[Robert Ardill]

Quote "I would perhaps think that a natural colour space may be XYZ…
https://www.flickr.com/photos/baxter43/5355713456/in/album-72157625678052039/On a serious note, I would say that working in a linear space is essential in order of keeping tonality changes consistent with colour, that is a change in density should not introduce a change in colour." end quote.


I have worked with several different raw converters and each and every one including ACR / LR; SilkPix; Capture One; DxO Labs; Olympus Viewer (my cameras software); DC Raw etc all produce different colour renditions. i.e. do a test shot with a GreyTag Macbeth Colour target and do a WB on the neutral patch then check the other colour patches and you will not see that none match. i.e. each and every raw conversion software applies their own colour profile (recipe).

Which is pretty bad in this day of nearly end-to-end color management.  It should be possible to photograph a color checker, say, in defined, controlled lighting conditions with a specified lens, ISO, raw ... and get pretty much identical results from all raw converters.  At least then we would be starting from a known point.

I can understand that for JPEGs that the conversion should reasonably aim for a pleasing rather than an accurate look, but surely one of the advantages of raw should be that we can get more color accuracy.

The XRite Color Checker Passport and Adobe DNG Profile Editor do a fairly decent job (but on a very small target, so not very accurate).  One could use Argyll CMS for example to create a proper icc file - but then one would have to use a raw converter that supports icc profiles.  I'm not sure it's worth the bother, but I haven't tried it ... it would be interesting to get some feedback from someone who has.

Robert

[Bart_van_der_Wolf]

Which is pretty bad in this day of nearly end-to-end color management.  It should be possible to photograph a color checker, say, in defined, controlled lighting conditions with a specified lens, ISO, raw ... and get pretty much identical results from all raw converters.  At least then we would be starting from a known point.

Hi Robert,

And that would be possible if the profiles were made for 'accurate' color, which they aren't in general. Profiles are a compromise, because the camera sensor CFAs see color different from how our eyes see color. In that translation/compromise, usually a kind of 'look' is used by the various producers, and that produces generally pleasing colors with somewhat smooth transitions, but not colorimetrically accurate.

I can understand that for JPEGs that the conversion should reasonably aim for a pleasing rather than an accurate look, but surely one of the advantages of raw should be that we can get more color accuracy.

And we can, but we then need to create specific profiles for specific goals.

Cheers,
Bart

[Robert Ardill]

Hi Robert,

And that would be possible if the profiles were made for 'accurate' color, which they aren't in general. Profiles are a compromise, because the camera sensor CFAs see color different from how our eyes see color.

Isn't that the whole idea behind color management, to translate between different devices and their foibles and to end up with an image that looks right to our eyes, on screen or printer ... and which is colorimetrically correct if we choose a colorimetric profile?

I suppose one thing we could do with raw converters that do not support icc profiles is to make an icc profile and assign it to the image once in TIFF.  But that would mean that all color adjustments would have to be done post raw.

Cheers,

Robert

[Robert Ardill]

Out of interest I made a matrix camera profile using ArgyllCMS and used this in DxO Optics Pro with very good results.  Admittedly this was based on the 24-patch colorchecker, so it can't be very accurate.  Still, here's the colorchecker with the profile applied:



and here is an image processed using C1 with the Argyll profile:



Certainly every bit as good as Lightroom with a DNG profile.

I basically used the process here (simplified): SteveHuff-Camera-Profiling

with these commands:

scanin -v -a -G1 -dipn A7RIICC.tif ref\ColorChecker.cht  ref\ColorChecker.cie
colprof -v -A“Sony” -M“A7RII” -D“Sony A7RII Overcast Matrix” -qm -am -nc -U1.47 A7RIICC

So this makes either C1 or DxO much more interesting for me.

Robert

[Robert Ardill]

In an Adobe raw workflow, doesn't matter! Use ProPhoto RGB:

sRGB urban legend Part 1


In this 30 minute video I'll cover:
Is there benefit or harm using a wider color gamut working space than the image data?
Should sRGB image data always be encoded into sRGB?
What are RGB working spaces and how do they differ?
What is Delta-E and how we use it to evaluate color differences.
Color Accuracy: what it really means, how we measure it!
Using Photoshop to numerically and visually see color differences when using differing working spaces.
Using ColorThink Pro and BableColor CT&A to show the effects of differing working space on our data and analyzing if using a smaller gamut working space is necessary.
Appendix: testing methodology, how differing raw converters encode into working spaces, capturing in-camera JPEG data and color accuracy.


Low resolution (YouTube): https://www.youtube.com/watch?v=1w0zUIl-dzY
High resoution: http://digitaldog.net/files/sRGBMyths.mp4

Hi Andrew.  As usual, a nice video and useful if you are someone who believes whoever it is who said that colors in one color space are colorimetrically more accurate than in another. 

Of course that isn't the issue at all ... it is because a wider gamut, given the same image bit-depth, will have greater interpolation errors.  That's a simple mathematical fact. So if the image gamut fits into a smaller color space, then all things being equal it's preferable to use the smaller color space. Admittedly this isn't so much of an issue with 16 bits.

Also, if your destination is for the web then you will sooner or later have to convert to sRGB. If you are working in a color space with a much larger gamut there is a very real risk of clipping (which can be adjusted for with soft-proofing, but the potential need for adjustment will always be present).  So if you know that your image is going to the web then you are better rendering it to sRGB than going to the wider-gamut color space first.  In addition, every time you convert an image from one color space to another you will introduce color shifts.

Again, if your monitor is 100% (or thereabouts) Adobe RGB, you are better being in Adobe RGB or sRGB as wider color spaces allow you to make adjustments to your image that are not visible (or rather, they will be automatically mapped to monitor RGB).

If your image has colors which are beyond Adobe RGB (easily seen by soft-proofing in Lightroom, for example) but are capable of being printed (again, you can see this with soft-proofing), well then by all means use a larger color space like Beta RGB or ProPhoto.  But then you do need to know what you are doing to avoid (potentially) ugly results on print.

So it's a question of using the right tools for the job.  If the desire is to simplify the workflow and to minimize the risk of artifacts then Adobe RGB is probably the best choice (at this point in time) for most of us with our Adobe RGB monitors.  If our images are always going to go to the web and not to print then we would be well advised to stick to sRGB.  If we typically have quite saturated images and wish to get the most saturated colors from our printer, well then a working space like ProPhoto might be the best choice for us.  If, like me, you're quite happy to process images differently for different output ... well then use all three.

Cheers,

Robert

[AlterEgo]

Out of interest I made a matrix camera profile using ArgyllCMS and used this in DxO Optics Pro with very good results.  Admittedly this was based on the 24-patch colorchecker, so it can't be very accurate. 

it is not like LUT profile, so 1000 patches will not give a magnitude better quality level vs a properly done process with CC Classic Mini.

[Robert Ardill]

it is not like LUT profile, so 1000 patches will not give a magnitude better quality level vs a properly done process with CC Classic Mini.

No, that's not correct (although for sure a badly done profile with lots of patches may well be worse than a well done profile with few patches).  24 patches just doesn't have enough colors (only two blues, for example).  It's just OK because digital cameras are essentially linear (as far as I know, that is).

[tho_mas]

No, that's not correct (although for sure a badly done profile with lots of patches may well be worse than a well done profile with few patches).  24 patches just doesn't have enough colors (only two blues, for example).  It's just OK because digital cameras are essentially linear (as far as I know, that is).

"Alter Ego's" post was not so much about the number of patches but primarily about matrix based vs. table based profiles ...

Again, if your monitor is 100% (or thereabouts) Adobe RGB, you are better being in Adobe RGB or sRGB as wider color spaces allow you to make adjustments to your image that are not visible (or rather, they will be automatically mapped to monitor RGB).

when the monitor gamut is the reason to choose a color space for editing... why don't you just use your monitor profile as working space?

[digitaldog]

Of course that isn't the issue at all ... it is because a wider gamut, given the same image bit-depth, will have greater interpolation errors.  That's a simple mathematical fact.

Fortunately the math also illustrates the differences are not visible.

[AlterEgo]

24 patches just doesn't have enough colors (only two blues, for example).  It's just OK because digital cameras are essentially linear (as far as I know, that is).

sure, if you want to start accounting for various nonlinear possibilities then even 3 x 1D TRCs, per channel (in matrix profiles) will not be enough (as you deal with raw data after demosaicking) and you will need 3D LUT to account fully... so no, you will not get a magnitude better matrix profile with CCSG or CCDC or custom made target vs CC24 for 3x3 matrix and if you want to account for exposure-based non linearities in you might as well try do bracketing when shooting CC24  but you still get restricted by just 3x3 matrix part that stays the same... too few degrees of freedom

[tho_mas]

Fortunately the math also illustrates the differences are not visible.

in your particular test, yes. Might be different if you did the same experiment with high saturated colors in a large color space. Who knows...
The "higher precission" of smaller color spaces comes into play when editing... IMHO. In a smaller color space you can make more subtle edits. On the other hand I have yet to see any limitations when editing in a larger color space in 16bit. So the debate whether larger color spaces are less precise or not is pretty academic to me personally... Much more important is workflow-discipline when working with large color spaces (especially those containing imaginary colors) - if you use a certain color space as a gamut warning profile when editing (PRMG, PhotogmutRGB or, if you whish, sRGB) nothing can go really wrong... IMHO...

[Robert Ardill]

it is not like LUT profile, so 1000 patches will not give a magnitude better quality level vs a properly done process with CC Classic Mini.

Have you actually tried this?  You may be right, but I would have thought that the CC Mini just doesn't have enough colors to give the 3 primaries accurately .. or to compute the TRC accurately (unless it is valid to assume that the camera response is entirely linear, which this article suggests that it is not Camera Linearity).

[Robert Ardill]

when the monitor gamut is the reason to choose a color space for editing... why don't you just use your monitor profile as working space?

Because my monitor is not linear and uses a LUT - a working space like ProPhoto or Adobe RGB should give me smoother gradients.

[Robert Ardill]

Fortunately the math also illustrates the differences are not visible.

True, providing we work in 16 bits, of course.  It's also true that for photos that are not highly saturated that the differences between sRGB, say, and ProPhoto will not be visible either.

Do you see an advantage in working in ProPhoto because it uses a D50 white point?  I would have thought so as there would be no need to transform from the raw converter white point (in most cases D50, I guess).  Which reminds me that this was one of the reasons I used Beta RGB: it has a D50 white point and is big enough for most photos without being over-bloated like ProPhoto.

The choice to me is that if I had to pick one and only one working space then I would most likely pick Beta RGB (or possibly ProPhoto as it's more widely used).  But I don't have to pick one working space and so if I'm developing a photo for the Web, say, I think it makes sense to use sRGB.  For print I will go back to Beta RGB (although I have been using Adobe RGB recently, mainly because it avoids me having to worry about colors that are not visible on my monitor)

[AlterEgo]

Have you actually tried this?  You may be right, but I would have thought that the CC Mini just doesn't have enough colors to give the 3 primaries accurately

with matrix profile you can't get a better precision in 3x3 matrix just by throwing more patches, more so by throwing more patches of the similar hues... you add more patches you change 3x3 matrix and as a result for some colors dE-whatever metric will get better and for some will actually get worse ... so you want to select actually just few important colors and find a good compromise between errors when building the matrix...

[digitaldog]

in your particular test, yes. Might be different if you did the same experiment with high saturated colors in a large color space. Who knows...

Might be different, doubtful considering the dE report of the thousands upon thousands of pixel (device values) who's dE differences in using either end of the working space gamut spectrum (sRGB vs. ProPhoto RGB) are tiny and invisible. Of course, you're encouraged to provide example images and dE reports that show otherwise.

[digitaldog]

True, providing we work in 16 bits, of course. 

That's the only way I fly and the only way my raw converters provide data from raw when correctly setup for rendering the data!

It's also true that for photos that are not highly saturated that the differences between sRGB, say, and ProPhoto will not be visible either.

Exactly, so just stick with ProPhoto RGB! It's no worse and often far better WHEN the data exceeds sRGB (which is pretty often in these parts).