Export from ProPhoto RGB to sRGB IEC61966-2.1?

[earlybird]

In another recent thread the discussion was beginning to hint at the implications of choosing among differing versions of sRGB as well as choosing rendering intents.

I've been trying to adopt a ProPhoto RGB workflow within Photoshop after working in sRGB for a long while. Almost everything I export for final output is in sRGB sRGB IEC61966-2.1 so it seemed like it made sense to use this workflow.

Using ProPhoto RGB in Photoshop seems to be working ok, but after reading some of the ideas expressed about converting from ProPhoto RGB to sRGB I am realizing that I am not very aware of the choices I might want to consider when making the color space conversion during my final output export.

My digital output has been primarily used for web and broadcast display, but lately I have become preoccupied with sending off for C-print and print on metal printing services. The printing services I use request sRGB colorspace and I have been enjoying the prints. I think they seem nicer than the prints I made when I printed my own stuff in a wet color lab way back when. I am satisfied but I also want to learn about what I do not know about.

What should I be thinking about when making my final conversion from 16bit ProPhoto RGB files to 8 bit sRGB jpeg files?

Thank you.

[digitaldog]

What should I be thinking about when making my final conversion from 16bit ProPhoto RGB files to 8 bit sRGB jpeg files?

Probably nothing more than selecting the profile and then converting to 8-bit per color. If it ain't broke...

[earlybird]

I was off reading about rendering intent choices. I wasn't much less confused about which choice to make after I read about it.

Relative Colormetric or Perceptual for photography of landscapes and nature subjects?

[Doug Gray]

I was off reading about rendering intent choices. I wasn't much less confused about which choice to make after I read about it.

Relative Colormetric or Perceptual for photography of landscapes and nature subjects?

Rendering intent comes into play when you print. Generally Perceptual produces more pleasing prints as it slightly increases contrast and may map colors that are outside the printer gamut into what the printer can print in a more attractive way. Relative intent doesn't change the contrast at all and, if you select black point correction can actually decrease contrast slightly and create a somewhat washed out appearance. Perceptual intent automatically does black point correction and by adding a slight increase in contrast tends to make the print look more like it does on a screen.

For RGB working space conversions there is no difference between Relative and Perceptual intent. At least for V2 profiles which are the ones in wide use including Photoshop, LR, etc.

[digitaldog]

Relative Colormetric or Perceptual for photography of landscapes and nature subjects?

Yes 
Which ever looks best to you. ICC profiles don't know anything about color in context.

[digitaldog]

]For RGB working space conversions there is no difference between Relative and Perceptual intent.

Yes, but to be clear, there's no Perceptual table, the software should know this and gray out Perceptual since it's not possible. Unfortunately we see two RI's and can only get one.

[Doug Gray]

Yes, but to be clear, there's no Perceptual table, the software should know this and gray out Perceptual since it's not possible. Unfortunately we see two RI's and can only get one.

Yep. Confuses a lot of folks. No Saturation or Absolute table either. Just the same matrix with the same values. Except for white point adjusting in Absolute which depends on the color engine. If you use the default, Adobe, there is no difference there either. OTOH, with Microsoft's engine there is a large difference in Absolute if the white points of the RGB space differ. Stick with the Adobe default. It's the correct interpretation of the ICC spec.

Enough to drive you crazy.

[xpatUSA]

Yep. Confuses a lot of folks. No Saturation or Absolute table either. Just the same matrix with the same values. Except for white point adjusting in Absolute which depends on the color engine. If you use the default, Adobe, there is no difference there either. OTOH, with Microsoft's engine there is a large difference in Absolute if the white points of the RGB space differ. Stick with the Adobe default. It's the correct interpretation of the ICC spec.

Enough to drive you crazy.

Some time back I was looking for a 'monitor' class sRGB profile with the tables in it as opposed to matrices. Purpose was to post saturated flower captures without sRGB gamut-boundary clipping. I actually found one but, after 3 computer deaths, it's not on the HD anymore and I can't remember where it's at

[Doug Gray]

Some time back I was looking for a 'monitor' class sRGB profile with the tables in it as opposed to matrices. Purpose was to post saturated flower captures without sRGB gamut-boundary clipping. I actually found one but, after 3 computer deaths, it's not on the HD anymore and I can't remember where it's at

You might try here:
http://color.org/srgbprofiles.xalter

[earlybird]

Thank you for the explanation about the use of rendering intents.

I am trying to condense the suggestions made in a few of the posts; I think what has been explained is that it is a practical concern when converting from a RGB color space to something like a CMYK color space, and if the print service I use requests sRGB files it is the print service provider's choice of rendering intent, made when converting the sRGB transport file into an actual printing profile color space, that has the most influence on the final results.

Does that seem correct?

[Tim Lookingbill]

Just FYI but I thought you'ld like to see (screenshot below tagged with my display profile) what some colors that were created in ProPhotoRGB on my calibrated system look like when converted to sRGB. I'm using an sRGB-ish LED display. Once they change like that you can't get them back by converting back to ProPhotoRGB.

Create the same colors on the left in Photoshop in a new document that is in (tagged or converted) to ProPhotoRGB using the color picker and then convert to sRGB to see if you get the same shift. Lab readouts between the two change by as much as ten points.

You may need to drag and drop onto your desktop and view in Photoshop because I've noticed various browsers show this differently.

[digitaldog]

You may need to drag and drop onto your desktop and view in Photoshop because I've noticed various browsers show this differently.

Safari here seems to do it fine. I see tiny differences.

[Doug Gray]

Just FYI but I thought you'ld like to see (screenshot below tagged with my display profile) what some colors that were created in ProPhotoRGB on my calibrated system look like when converted to sRGB. I'm using an sRGB-ish LED display. Once they change like that you can't get them back by converting back to ProPhotoRGB.

Create the same colors on the left in Photoshop in a new document that is in (tagged or converted) to ProPhotoRGB using the color picker and then convert to sRGB to see if you get the same shift. Lab readouts between the two change by as much as ten points.

You may need to drag and drop onto your desktop and view in Photoshop because I've noticed various browsers show this differently.

Assuming colors in ProPhoto outside of sRGB and the monitor gamut, the process used in displaying ProPhoto RGB colors is the same as converting to sRGB except that for displaying, the colors are converted to the monitor's profile instead of sRGB. The closer the monitor is to sRGB, the less change one will see converting to sRGB.

The largest change in colors will be seen by people that have profiled, wide gamut monitors.

When colors are converted from one RGB colorspace to another and both are matrix based, the math first converts to gamma=1, then a matrix multiply is used which adjusts for possible Bradford adaption together with the different RGB xyY values. Then the gamma is reapplied to whatever the target gamma is. Finally, the values are clipped. For 8 bit systems this is 0 and 255. Until the values are clipped the colors are unchanged. You can, for instance, have sRGB values that include the entire ProPhoto gamut if you were to allow the sRGB values to go over 255 and below 0. clipping is what changes the colors when you convert to a smaller space.

[Tim Lookingbill]

The largest change in colors will be seen by people that have profiled, wide gamut monitors.

Assuming someone here has a wide gamut profiled monitor do you get color shifts of similarly created colors when converting to sRGB? Is it worse than what I've posted?

I have some Raw images of sunsets fluorescing back lit leaves and grass editing to bring out the brilliance and richness in ProPhotoRGB in ACR and upon opening in Photoshop and converting to sRGB notice a reduction in saturation and a slight hue change toward cyan in the greens. To reduce this I have to convert to my sRGB-ish monitor profile when posting online.

I always thought it was an ICC transform display bug and that no one would see it except those viewing on my display, but if it's as you've technically described, Doug, then it's permanent when converting to sRGB and everyone does see this. There's no way around this. Is this correct?

[Tim Lookingbill]

Another RGB color space encoding oddity is that png I posted opened in Photoshop in my monitor space since I didn't do any conversion after capture has the bottom green at...

ProPhoto: 128,225,0/Lab: L81,a-50,b90 and
sRGB: 128,225,58/Lab: L81,a-49,b77 (that's a 14 point difference in Lab space)

Convert to ProPhoto...

ProPhoto: 146,205,52/Lab same as above
sRGB: 147,206,68/Lab same as above

None of the previews changed after doing the conversion which proves it's permanent in regards to previews when converting to sRGB.

[Doug Gray]

Another RGB color space encoding oddity is that png I posted opened in Photoshop in my monitor space since I didn't do any conversion after capture has the bottom green at...

ProPhoto: 128,225,0/Lab: L81,a-50,b90 and
sRGB: 128,225,58/Lab: L81,a-49,b77 (that's a 14 point difference in Lab space)

Convert to ProPhoto...

ProPhoto: 146,205,52/Lab same as above
sRGB: 147,206,68/Lab same as above

None of the previews changed after doing the conversion which proves it's permanent in regards to previews when converting to sRGB.

I have a CG318. When at native (wide gamut) the right and left sides are distinguishable. When I switch to sRGB, the two sides look exactly the same.

Precisely what I would expect.

BTW, the attached profile is only slightly larger than sRGB. The biggest difference is the Green which has a xy coordinate of .3349, .6150

sRGB's G is at .3212, .5978

For comparison, Adobe RGB's G is at .2302, .7016 which is considerably more green.

[Tim Lookingbill]

Try this test if you don't mind, Doug, on your wide gamut monitor.

In a New Document in ProPhotoRGB in Photoshop create an RGB fill of 146,205,52 on a white background and convert to sRGB.

Do you see as pronounced a shift as what I've posted in the screenshot viewed on your wide gamut space display?

I mean if you're seeing more of a shift that means some folks including myself who edit on sRGB-ish displays may not realize that others viewing on profiled wide gamut displays aren't seeing the colors intended posted online. Just would like to get an idea of the severity of the shift. Of course not all colors viewed as flat fill swatches shift just as it is in most photos, but if editing sunsets and other saturated or intensely colored scenes it may not look right to some degree without the photographer ever realizing it.

So would you say this shift by converting to sRGB permanently changes the images where further editing in that space will most likely make it worse?

[Doug Gray]

Try this test if you don't mind, Doug, on your wide gamut monitor.

In a New Document in ProPhotoRGB in Photoshop create an RGB fill of 146,205,52 on a white background and convert to sRGB.

Do you see as pronounced a shift as what I've posted in the screenshot viewed on your wide gamut space display?

I mean if you're seeing more of a shift that means some folks including myself who edit on sRGB-ish displays may not realize that others viewing on profiled wide gamut displays aren't seeing the colors intended posted online. Just would like to get an idea of the severity of the shift. Of course not all colors viewed as flat fill swatches shift just as it is in most photos, but if editing sunsets and other saturated or intensely colored scenes it may not look right to some degree without the photographer ever realizing it.

So would you say this shift by converting to sRGB permanently changes the images where further editing in that space will most likely make it worse?

Yes, I see a visible shift. It's not that strong even though the Lab values are quite different with a DeltaE76 of 14. The perceptual deltaE2000 is only 3.4.
I measured the colors and the monitor is within a dE2k of .3


For what it's worth if I create a RGB (0,255,0) patch using my monitor's native colorspace I get a huge color shift.  In ProPhoto color is (70,243,57) (Lab=86,-128,95). When converted to sRGB it is still (0,255,0). When converted back from sRGB to ProPhoto it is now (138,237,78) (Lab=88,-79,81) which is very desaturated.

If I view both patches with the monitor set to sRGB mode they look identical. With the monitor set in native (wide) gamut the patches are very different.


I would like to this additional note.

The fact you are seeing any color shift at all is because your monitor's "green" is slightly more green than an sRGB green. I know this because I extracted and examined the monitor profile attached to your image. As big as that seems to you it is much worse with wide gamut monitors. Folks that use wide gamut monitors have to be extremely careful when converting to sRGB as there can be very big color shifts. One work around is to just use the proof view settings using sRGB as the target. Then, WYSIWYG when converting to sRGB.

[xpatUSA]

You might try here:
http://color.org/srgbprofiles.xalter

Oops - should have said I was looking a V2 profile, not V4, :-(

Sorry.

[Tim Lookingbill]

Yes, I see a visible shift. It's not that strong even though the Lab values are quite different with a DeltaE76 of 14. The perceptual deltaE2000 is only 3.4.
I measured the colors and the monitor is within a dE2k of .3


For what it's worth if I create a RGB (0,255,0) patch using my monitor's native colorspace I get a huge color shift.  In ProPhoto color is (70,243,57) (Lab=86,-128,95). When converted to sRGB it is still (0,255,0). When converted back from sRGB to ProPhoto it is now (138,237,78) (Lab=88,-79,81) which is very desaturated.

If I view both patches with the monitor set to sRGB mode they look identical. With the monitor set in native (wide) gamut the patches are very different.


I would like to this additional note.

The fact you are seeing any color shift at all is because your monitor's "green" is slightly more green than an sRGB green. I know this because I extracted and examined the monitor profile attached to your image. As big as that seems to you it is much worse with wide gamut monitors. Folks that use wide gamut monitors have to be extremely careful when converting to sRGB as there can be very big color shifts. One work around is to just use the proof view settings using sRGB as the target. Then, WYSIWYG when converting to sRGB.

Thanks, Doug, for taking the time to conduct my little test in addition to your other tests, Delta E information and wide gamut display issues converting to sRGB. 

It does make sense about the green channel because I've noticed slightly increased overall definition (and change in luminance) shifting the green/magenta Tint slider in ACR/LR more toward magenta. A lot of sunsets require this to get the golden yellow because camera profiles just can't get there. It gets pretty ugly having a greenish 2800K amber sunset tinting all the trees, rocks and foliage. Adobe Standard sometimes does a better job over a custom DNG profile.