ProPhoto for RAW-Output

[Dieter Mayr]

I recently got a new printer and found out that it's Gamut with my favourite paper exceeds my standard working CS Chrome2000 D65 as well as WideGamutRGB, especially in intense yellow tones.
So, I wanted to make some tests with ProPhotRGB, but in the current version PWP does not support ProPhoto as output CS of the RAW-conversion.
DCRaw supports it ( http://www.cybercom.net/~dcoffin/dcraw/dcraw.1.html ), and directly developing some raws with intense yellows showed that there are colors outside even WideGamut that my printer could print.
Are there chances to get ProPhoto as a output profile for RAWs in PWP too ?
I'm aware that ProPhoto is not the CS for evertyday use, especially when working with 8 bit or JPEGs, but for special cases it seems to be optimal.

[Charles2]

As you note, for such rare exceptions where demosaicing into ProPhotoRGB makes sense, you can start with something like dcraw or Raw Therapee. I ask out of curiosity, how do you find that an image has intense yellows beyond Chrome 2000 D65?

[afx]

Hmm,
as I do not use PWP as a raw converter but only for some retouching work I am a bit puzzled here.
Why would anyone use anything else but ProPhoto (or linear variants of it) for handling the output of the raw conversion?
DSLRs have such a wide gamut that most of the other profiles are to small to handle it. After all there is a reason why all the serious raw converters use variants of ProPhoto as their internal working space.

cheers
afx

[Dieter Mayr]

Charles, I had the chance to run a test with ColorThink at a friends place.
One can plot colorspaces, and also load pictures to show their actual gamut.

Gamut_Sunflower_s.jpg (16.51 KiB) Viewed 7295 times

The solid is Chrome2000D65 and the cloud of points is the image.
The image is one of a sunflower, light from back, developed to ProPhoto.

Sunflower_s.jpg (29.12 KiB) Viewed 7298 times

The image here is in sRGB for better appearence on the web.

[jsachs]

I think you may be misinterpreting what is happening with raw files and camera profiles. Here is my understanding:

Digital camera sensors simply produce a set of RGB values which are not referenced to any particular primaries. Determining the actual camera color gamut is actually very difficult as it cannot be measured directly (there are some papers on this on the web). Camera manufacturers generally try to produce RGB values compatible with sRGB or Adobe RGB so it makes sense to use whichever of these color spaces you specified in your camera settings unless you have a custom profile created specifically for your camera. Simply tagging the RGB data with a wider color space does not recover any otherwise clipped color data - it just reinterprets the raw RGB values as being from an arbitrarily wider gamut than normal. When the raw converter in turn converts the tagged image to the working color space, it uses perceptual color mapping so the image gamut is simply shrunk to fit the working color space gamut unless the working color space is also a wide color space. In other words, using a wide color space as a camera profile is really a useless exercise that in the end is at best more or less equivalent to a saturation boost.

[Charles2]

Camera manufacturers generally try to produce RGB values compatible with sRGB or Adobe RGB so it makes sense to use whichever of these color spaces you specified in your camera settings

Thanks for some light on a mystery I wondered about.

Most enthusiast and professional cameras have a choice of sRGB and Adobe RGB. I wonder whether the raw file differs depending on the choice. It's not often documented, and I'd guess that in many cases the selected space is used only to generate the camera JPG.

The Fuji X-Trans cameras hint at the situation. When you shoot raw + JPG, you can generate additional JPGs in playback mode in the camera, varying white balance, saturation, and so on -- including color space. This behavior suggests that the raw is just there, in a mysterious profile. I suppose that when one uses the supplied Fuji Raw File Converter program in one's computer, it too knows how to make an sRGB or Adobe RGB file as the one chooses.

[ksinkel]

Dieter,

I am not sure if you need to translate your image to a wide gamut at the time you convert from raw. I wonder if it wouldn't be possible to make the gamut expansion just before printing, instead.

Here is my reasoning:

You are not primarily interested in accurate color but in the best print. You are translating an outdoor scene which includes backlighting to a print which will be viewed by reflective light. I think it's pretty much accepted that the print cannot come close to matching the dynamic range of the original scene. So the typical way the impact of the original high dynamic range is preserved is by expanding the gamut of the print, using more intense colors in place of brighter ones. This is similar to how b+w photographers using the zone system manipulated tonality for maximum impact. They disregarded the contrast of the original scene and concentrated on filling the entire tonality range of their materials. In effect I am suggesting doing the same but for color rather than tonality.

Thus it seems to me that after adjusting your print for maximum impact on the screen (which has a narrower gamut but greater dynamic range than your printer), you would want to expand the gamut by converting the color profile just before printing.

I would be interested to know how this works in practice.

Kiril

[Dieter Mayr]

Jonathan, Kiril, thanks for your explanations.
My understanding of colorspace and RAW files was that one should use a "as big as possible" colorspace, i.e. ProPhoto to be able to contain the full gamut of the camera, and then of course the same huge colorspace as woriking colorspace, and of course 16 bit because of the risk of banding in 8 bit with such a huge cs.
Scaling down to a smaller colorspace, say sRGB for web output or a "drugstore photostation" can always be done, but converting to a bigger colorspace should be avoided for quality reasons.
So far my understanding from what I have read.
But with your thoughts in mind, it seems logical that one can shrink or expand colorspaces with Maintain Full Gamut = Perceptual rendering intent without losses as long the bit count stays the same (16 bit in our case).
I will have to think deeper about this whole thing and do some practice tests but yes, seems logical to me.

[afx]

My understanding of colorspace and RAW files was that one should use a "as big as possible" colorspace, i.e. ProPhoto to be able to contain the full gamut of the camera, and then of course the same huge colorspace as woriking colorspace, and of course 16 bit because of the risk of banding in 8 bit with such a huge cs.

Easy to verify...
See http://www.afximages.com/articles.php?a ... lorProfile

Scaling down to a smaller colorspace, say sRGB for web output or a "drugstore photostation" can always be done, but converting to a bigger colorspace should be avoided for quality reasons.

Converting from small to big will give you further headroom for adjustments, but any information that was lost by going small initially will be lost forever.

But with your thoughts in mind, it seems logical that one can shrink or expand colorspaces with Maintain Full Gamut = Perceptual rendering intent without losses as long the bit count stays the same (16 bit in our case).

That does not work accurately. Visually it might work in many cases, but I consider it a gamble.
Every time you go from big to small, you risk loosing detail information because you have a transformation that can not guaranty all information is included, with perceptual rendering you get a best match only...

I will have to think deeper about this whole thing and do some practice tests but yes, seems logical to me.

I've made an experiment with images of a red rose in my garden and then compared prints that went through several different color spaces. See the referenced link. In the end, you want to stay wide throughout the whole pipeline....

I think a lot of confusion is generated by the in camera working color space settings. They are irrelevant for raw files.
In the process of the raw file decoding, scene referenced image data is converted to a standard output referred image with a defined working space. And that working space should be wide enough to encompass whatever the camera delivers.

cheers
afx