10 bit RAW converters and editors

[digitaldog]

Oh, that. I thought everybody knew that . .

That and 15 minutes can save you 15 percent or more on car insurance.

[Lundberg02]

Now is everyone jolly again and I can go back to my nap?

[xpatUSA]

Now is everyone jolly again and I can go back to my nap?

Sure. While napping, let your mind work on: RawTherapee has some .icc color profiles with look-up tables, not matrices 

[Lundberg02]

I have RT, I'll give their rendering intents a chance, despite the geekiness.

[digitaldog]

I have RT, I'll give their rendering intents a chance, despite the geekiness.

Don't expect it will be the best thing since sliced bread. No farce  .

[Lundberg02]

Can they found somewhere in RT and copied out to be installed in Photoshop? I would think so if they're actual icc profiles.

[xpatUSA]

Can they found somewhere in RT and copied out to be installed in Photoshop? I would think so if they're actual icc profiles.

On my computer, they are located here:

C:\Program Files\RawTherapee-4.0.11.203\iccprofiles\output\

[Lundberg02]

They're probably in Application Support on my Mac.   Oooops, no, not in either my HD Library or my username Library.  Hm.
OK, right click on the app gives you package contents. Several profiles folders with sub folders. The icc profiles folder has a couple called large, one medium and several sRGB. There are no profiles identifiable as Pro Photo or Adobe RGB in any profile folder or sub folder, but there must be fifty total.

[xpatUSA]

They're probably in Application Support on my Mac.   Oooops, no, not in either my HD Library or my username Library.  Hm.
OK, right click on the app gives you package contents. Several profiles folders with sub folders. The icc profiles folder has a couple called large, one medium and several sRGB [sounds familiar]. There are no profiles identifiable as Pro Photo or Adobe RGB in any profile folder or sub folder, but there must be fifty total.

My old Mac is still on System 9.0 and haven't touched it for a while, so I have no idea what you're talking about.

"identifiable" . . Hm . . .  er, did your RawTherapee come with a manual by any chance?  

Here's what it says:

"RawTherapee comes bundled with a number of custom-made high quality
output profiles:
■ RT_sRGB similar to sRGB with gamma close to sRGB: g=2.40,
slope=12.92
■ RT_sRGB_gBT709 similar to sRGB with gamma BT709: g=2.22,
slope=4.5
■ RT_sRGB_g10 similar to sRGB with linear gamma g=1.0, slope=0
■ RT_Middle_gsRGB similar to AdobeRGB1998 with gamma close to
sRGB: g=2.40, slope=12.92
■ RT_Large_gsRGB similar to ProPhoto with gamma close to sRGB
g=2.40, slope=12.92 (close to "Melissa" used by Lightroom)
■ RT_Large_gBT709 similar to ProPhoto with gamma BT709: g=2.22,
slop=4.5
■ RT_Large_g10 similar to ProPhoto with linear gammma g=1.0,
slope=0"

and here's my folder like yours, which appears to contain all of the above:



good luck . . .

[Lundberg02]

Thanks for that.   I have the manual somewhere, I have tried RT but was put off by its dark UI and inability to find my RAWs.  I had thought that the Large and Middle probably corresponded to the conventional naming and the sRGB was obvious. I don't know why geeks insist on obfuscating everything.
OK I just read the Rendering Intents section, I get it.  I also see that I have to tell RT what my monitor profile is.

[digitaldog]

"RawTherapee comes bundled with a number of custom-made high quality
output profiles:

None of which are output profiles but whatever. The question is however, are they LUT based and have all three RI tables? If not, nothing you could not have built yourself inside of Photoshop. Super easy to take ProPhoto and build a 1.0 TRC ICC profile.

[xpatUSA]

I don't know why geeks insist on obfuscating everything.

Photography is one of the most obfuscatory disciplines on the planet, IMHO.

A caveat as regards RawTherapee's 'ProPhoto' .icc (.icm actually but makes no difference) files:

The non-linear parameters are incorrect. Since ProPhoto is actually Kodak's ROMM re-named, the gamma should be 1.8 with a 'shadow' linear slope of 16.

Not certain what @digitaldog means by "none of which are output profiles". If it looks like a duck, etc. 

As to LUT's I can only say that my system sRGB profile is 3.07Kb and RT_sRGB is 24.9Kb which indicates more than just 'matrix-based'. Can't be sure though.

cheers,

[digitaldog]

Not certain what @digitaldog means by "none of which are output profiles". If it looks like a duck, etc. 

There are classes of ICC profiles. The profiles mentioned are not output profiles, that's all. They are working space profiles. If you want to call a duck a chicken because the feather's look the same to you, OK with me but it's incorrect. None of the RGB working space are based on any real output device and in fact are theatrically constructed. What output device can produce ProPhoto RGB?

[digitaldog]

As to LUT's I can only say that my system sRGB profile is 3.07Kb and RT_sRGB is 24.9Kb which indicates more than just 'matrix-based'. Can't be sure though.

If there's only the colorimetric table, big deal. You have those in Photoshop.

[xpatUSA]

There are classes of ICC profiles. The profiles mentioned are not output profiles, that's all. They are working space profiles.

Thank you. I'm beginning to get it. The profiles I mentioned are of type 'mntr'. They do not have A2B or B2A tables. As opposed to ICC V4 which do, and even then there would need to be A2B and B2A tables specifically for 'perceptual'. Does PhotoShop honor V4 profiles?

If you want to call a duck a chicken because the feather's look the same to you, OK with me but it's incorrect.

Har-de-har-de-har.

None of the RGB working space are based on any real output device and in fact are theatrically constructed.

What output device can produce ProPhoto RGB?

My home-crafted laser display with XYZ primaries: 0.7976685, 0.2880402, 0 : 0.1351929, 0.7118835, 0 : 0.0313416, 0.0000916, 0.8249054
 (haven't actually built it yet, having difficulty finding the diodes, can't image why).

If there's only the colorimetric table, big deal. You have those in Photoshop.

I don't have PhotoShop. No need for it ay my level 

Thank you for your ever-humble response. Now I have go eat crow with the OP  

[xpatUSA]

Thanks for that.   I have the manual somewhere, I have tried RT but was put off by its dark UI and inability to find my RAWs.  I had thought that the Large and Middle probably corresponded to the conventional naming and the sRGB was obvious. I don't know why geeks insist on obfuscating everything.
OK I just read the Rendering Intents section, I get it.  I also see that I have to tell RT what my monitor profile is.

Well, it's humble pie time for me. Another illusion shattered   

Looks like RawTherapee is not up to it. See my post above.

I'm going to experiment in RawTherapee with a ICC V4 profile. I'll post anything significant.

[digitaldog]

Thank you. I'm beginning to get it. The profiles I mentioned are of type 'mntr'. They do not have A2B or B2A tables. As opposed to ICC V4 which do, and even then there would need to be A2B and B2A tables specifically for 'perceptual'. Does PhotoShop honor V4 profiles?

mntr is a 'display profile' which makes sense as all RGB working spaces are based upon some theoretical display. Even sRGB. Or ProPhoto RGB (which has primaries outside human vision which makes this fun). Photoshop has no issues with properly built V4 profiles. But almost all V4 profiles are V2 in sheeps clothing. They usually don't support what is called the PRMG which is what makes those profiles interesting.

[xpatUSA]

I'm going to experiment in RawTherapee with a ICC V4 profile. I'll post anything significant.

I moved the V4 'display' icc profile into the RawTherapee (RT) icc output folder. I took my favorite highly saturated sunflower shot and exported it from Sigma Photo Pro as a TIFF with the SPP ProPhoto profile embedded. I opened it in RT with 'use embedded profile' selected and, in color management, set the output profile firstly to RT_sRGB and saved as a TIFF and then again but with the V4 profile.

Then I compared the two files in ColorThink. The 3D Lab gamuts were significantly different, so RT at least did something!

The V4 yellows were away from the sRGB boundary, RT_sRGB hard up against it.

The V4 hues were rotated anti-clockwise about 10 degs. However, the RT_sRGB had a bad dog-leg at the highest yellows also anti-clockwise.

The V4 darks were raised to approx L* = 3, but the RT_sRGB darks went down to zero L*.

They usually don't support what is called the PRMG

Had to Google PRMG: http://www.color.org/v4_prmg.xalter

In FastStone Viewer (FSV), there is some peculiar blue coloration in the dark areas of the rendered image - could that be PRMG in action? I would be surprised because FSV doesn't even embed profiles in saved images!

Not sure who "They" are but I downloaded the display-preferred V4 from color.org some time back and have not modified it.

[digitaldog]

Then I compared the two files in ColorThink. The 3D Lab gamuts were significantly different, so RT at least did something!

I'd expect different profiles to produce differences that you report. Which looks better? Which prints better? One of the 'issues' with V2 profile that V4 was supposed to fix was the assumed 'source' color space within a conversion. With V2, that's not defined like it can be in V4. Let's say you start with data in Adobe RGB (1998) and you want to convert it to an output color space (MyEpsonRGB). The conversion goes Adobe RGB (1998) to Lab, then Lab (the PCS) to Epson. But when in Lab, the CMS doesn't 'know' the original was Adobe RGB (1998), it only knows about the Lab data. Each profile maker has to come up with some assumed 'source' and they can differ so the results will differ too. V4 was supposed to introduce the PRMG which helps define all this but without a PRMG, we're back to V4 profiles in sheeps clothing. Plus I suspect you're still forced to use a Colorimetric RI for the conversions. Add a Perceptual table, plus the fact that there are no rules in how one build this conversion for Perceptual, you'll see differences converting the same original data through those profiles. Again, which is better?

[xpatUSA]

I'd expect different profiles to produce differences that you report. Which looks better? Which prints better? One of the 'issues' with V2 profile that V4 was supposed to fix was the assumed 'source' color space within a conversion. With V2, that's not defined like it can be in V4. Let's say you start with data in Adobe RGB (1998) and you want to convert it to an output color space (MyEpsonRGB). The conversion goes Adobe RGB (1998) to Lab, then Lab (the PCS) to Epson. But when in Lab, the CMS doesn't 'know' the original was Adobe RGB (1998), it only knows about the Lab data. Each profile maker has to come up with some assumed 'source' and they can differ so the results will differ too. V4 was supposed to introduce the PRMG which helps define all this but without a PRMG, we're back to V4 profiles in sheeps clothing.

Wow. I naively thought that the PCS is a perfect buffer, so to speak. Now I learn to my horror that something on output side has to 'know' what the input side was.

 

Plus I suspect you're still forced to use a Colorimetric RI for the conversions.

My V4 profile has tables for both Perceptual and Colorimetric, four tables in total. RawTherapee allows the output intent to be set as Perceptual. However, as you have helped me realize, setting it doesn't necessarily mean you get it 

Add a Perceptual table, plus the fact that there are no rules in how one build this conversion for Perceptual, you'll see differences converting the same original data through those profiles.

RawTherapee allows selection of the Input Profile (I use 'embedded'), Working Space (ProPhoto recommended default), and the Output Profile. Pardon me for going on about RT, but my other editor PSE6 is a bit limited from a CM POV.

Again, which is better?

I had tried V4 in the past and given up on it, using other means of turning highly saturated captures into decent-looking shots. Now I've learned more about it from your good self, I may play some more.

In answer to your questions - RT_sRGB looks 'better' than V4 out of the box on my screen. I don't print anything.