I like the last phrase: Nonetheless, camera input profiles do capture much larger percentages of real colors than can be contained within the sRGB color gamut.
Now that there exist UHDTV displays with REC.2020 color space, I wonder how to display these colors. I tried following approach:
In PS CameraRaw save DNG image as 16bit tiff with ProPhotoRGB color space.
In ImageMagick convert the tiff to color profile REC2020-elle-V???
On
I find about 12 profiles and I am not sure which one is correct.
I then further process the tiff with ffmpeg and create a X265 video.
One option would be to use our libre software, some of which supports Rec. 2020.
RawTherapee has a Rec. 2020 working and output profile - no ImageMagick required.
To explain how to read the profile names from your GitHub link we need to ask @Elle.
I am am aware of RawTherapee. I just use Camera Raw because I want to convert my “old” pictures. They all have a crop and other parameters (lens correction…) defined in PS CR. As far as I know I would have to define all this parameters again in RawTherapee. Maybe some day there will be a converter. Shouldn’t be too difficult to convert an XMP file.
Take all this with a big grain of salt, as I’m learning as I make mistakes…
The different profiles are basically 1) ICC Version 2 versus Version 4, and 2) different tone response curves, how the tones are scaled (or not).
For version, I guess I’d go with V2 as there’s a lot of discussion about V4 incompatibilities; you might try both V2 and V4 with your chosen display chain to see if either/both work.
TRC was a bit non-obvious, as @Elle states she produces profiles, “…with the profile’s standard Tone Reproduction Curve (“TRC”), the linear gamma, gamma 1.8, and gamma 2.2 TRCs, and the sRGB, LAB “L”, and Rec709 TRCs, in both V2 and V4 versions.” I couldn’t find a separate file for any with a nomenclature identifying the profile’s standard TRC, so I went looking for what it might be. Of course Wikipedia comes to the rescue, Rec. 2020 - Wikipedia, which states “that is the same nonlinear transfer function that is used by Rec. 709, except that its parameters are given with higher precision”
All that, I guess I’d go with Rec2020-elle-V2-rec709.icc, and maybe the V4 version for grins. From a decent working profile, you should be using software that lets you convert to a file profile on output. Raw Therapee will do that, with Camera Raw it looks like you maybe could just save the 16-bit TIFF with the Rec2020 profile, per your Step 1.
Hi @Morgan_Hardwood - My profile names hopefully should be easy to interpret. I followed a naming convention that is similar to something suggested - I think by Kai Uwe Behrmann a long time ago - as a way to deal with users being faced with multiple versions of nominally “the same” ICC profile color space supplied by multiple ICC profile vendors:
Base Color space: Rec2020-
Vendor: elle-
ICC profile spec version: V2- or V4-
the Profile TRC (tone reproduction curve), which for my profiles is one of these:
simple gamma curve TRC: g10/g18/g22 - meaning the TRC is respectively either gamma 1.0, gamma 1.8, or gamma 2.2.
a more complicated curve: srgbtrc/rec709/labl - meaning the TRC is respectively either the sRGB TRC, the Rec.709 TRC, or the LAB companding curve
the extension, which is either “.icc” or “.icm”. I prefer “.icc”.
So if, for example, RawTherapee ICC profile names followed this convention, then the RT LargeRGB with the gamma=1.80 TRC might be something like this, assuming the RT profiles are V4 rather than V2 (without checking, I think they might be V2?):
LargeRGB-RT-V4-g18.icc
Hi @george84 - I don’t know which, if any, of my Rec2020 profiles is the right profile for using with a X265 video. I don’t have the Rec2020 spec in front of me at the moment, though I’ve read a couple preliminary versions (the metadata in my Rec2020 profiles has a link, that’s hopefully still good). The funny thing about these specs is that once an official version is released, you have to buy it if you want to read it, and they are not cheap! But the “pre-release” specs can often be found online.
I remember seeing two different TRCs given in the Rec.2020 spec, and one of the TRCs is definitely the Rec.709 TRC. If this is the right TRC for your purpose, then use Rec2020-elle-V*-rec709.icc, where whether you should use the V2 or V4 version depends on whether the application that will display your image is capable of reading V4 profiles. Or use the Rec2020.icm from argyllcms.com - if you download the compiled code for your OS, it’s in the “rec” folder.
I see @ggbutcher has already recommended the rec709 version (he posted while I was still typing ) and pointed you to the Wikipedia article.
The only reason I hesitate to say "Yes, use the “-rec709.icc” version is because the Wikipedia article says something about a version of the profile for display purposes, using a gamma=2.4 TRC, and none of my ICC profiles have a gamma=2.4 TRC.
Also, I’m very unclear about using different TRCs for editing, vs displaying images on devices calibrated to various standards. I assume these new UHDTV displays aren’t using ICC profile color management but rather expect the signal to already be correct for the display. Hopefully someone else has more definitive information.
Thank you @Elle and @ggbutcher. I went with V4-Rec709 because X265 can use this trc as input. To come back to my original goal: To see the additional colors on TV. I find it very difficult to judge it by eye. My original Picture was a Pentax K-5 DNG which should be 14 bits, so I was hoping to see a difference compared to sRGB. I have the impression that the new colors are a little more saturated, the greens and oranges seem better. But this is not objective. Also the TV is not calibrated. If anyone is interested I can be more precise how the video is created from tiff.
Consider it to be next to impossible in general. We try to support Lr XMP in darktable but only manage to support basic stuff like cropping, rotation, exposure, vignette and some others. Full support for other raw converter’s parameters is too much depending on the internal workings that it’s not even feasible among open source programs, let alone closed ones where we have no way of understanding what the values do.
) and pointed you to the Wikipedia article.