Film simulations and wide gamut

I'm processing my backlog and ran into the rare cases of light sources with colors out of the usual sRGB space. darktable doesn't have an enhanced color matrix for the X-T20, and the standard color matrix results in color clipping badly. I had to switch to a larger linear Rec2020 RGB or linear prophoto RGB but of course the colors then are slightly off.

I fed the RAF file back to the X-T20 and generated the JPGs for different film sims. From the result, looks like Astia is the most accurate, still has plenty of saturation and contrast while staying true to the original colors. Velvia pushed the saturation a little too far, changing saturated blue into purple/violet.

I'm just curious how different RAW processors (LR, C1, ...) handle the wide gamut. Here are the two RAF files if anyone can give me a hand with your favorite tools. Thanks.


The SOOC JPGs generated by the X-T20:


Atsia SOOC




Velvia SOOC, bright blue discs turned into purple/violet

Here is a typical output in darktable using standard color matrix as color input profile. Even without applying any curve/filmic, the saturated blue discs are clipped into solid blue. The colors are mostly correct.


standard color matrix mapped into linear prophoto RGB

And here's one using linear prophoto RGB as color input profile. There's no clipping, the blue discs stay half-transparent, but the colors are slightly off due to incorrect mapping, they do look similar to SOOC ProNeg.


linear prophoto RGB mapped into linear prophoto RGB

Interestingly, there's a discussion on color spaces just last week: Color spaces nightmares: gamut clipping

.

sluggy_warrior said:

Here is a typical output in darktable using standard color matrix as color input profile.

Click to expand...

quick tip: you may try the color calibration module to pull colors into gamut.

Will try to show something later

His Dudeness said:

quick tip: you may try the color calibration module to pull colors into gamut.

Click to expand...

thanks, I've been using the color calibration module instead of the white balance module. Setting the right WB can often time put the gamut right back in, but sometimes I prefer a different (warmer) WB.

sluggy_warrior said:

I'm processing my backlog and ran into the rare cases of light sources with colors out of the usual sRGB space. darktable doesn't have an enhanced color matrix for the X-T20, and the standard color matrix results in color clipping badly. I had to switch to a larger linear Rec2020 RGB or linear prophoto RGB but of course the colors then are slightly off.

I fed the RAF file back to the X-T20 and generated the JPGs for different film sims. From the result, looks like Astia is the most accurate, still has plenty of saturation and contrast while staying true to the original colors. Velvia pushed the saturation a little too far, changing saturated blue into purple/violet.

I'm just curious how different RAW processors (LR, C1, ...) handle the wide gamut. Here are the two RAF files if anyone can give me a hand with your favorite tools. Thanks.

https://drive.google.com/drive/folders/1UH8tlVtrJUUR1f_vlSbMkCLWrzCCjS1x?usp=sharing

The SOOC JPGs generated by the X-T20:


Atsia SOOC


Velvia SOOC, bright blue discs turned into purple/violet

Click to expand...

I think this is just a clipping, not a gamut problem. here is just pulling back the blue channel back (mostly) from beyond the clipping threshold (Lightroom). Same profile and color settings.

Erik Baumgartner said:

I think this is just a clipping, not a gamut problem. here is just pulling back the blue channel back (mostly) from beyond the clipping threshold (Lightroom). Same profile and color settings.

Click to expand...

Thanks. It could just be darktable, being a FOSS tool, doesn't have the resources to profile most cameras, thus, its standard matrix is just not as good as the commercial tools.

In your screenshots, although most of the blue are pulled back, the bright green circles now look a little off (greyish)?

sluggy_warrior said:

His Dudeness said:

quick tip: you may try the color calibration module to pull colors into gamut.

Click to expand...

thanks, I've been using the color calibration module instead of the white balance module. Setting the right WB can often time put the gamut right back in, but sometimes I prefer a different (warmer) WB.

Click to expand...

color calibration is not only white balance. It is also a channel mixer.

Take a look at the tab "colorfulness" and the "gamut compression" in the "CAT" tab. I try to post an example later.

His Dudeness said:

color calibration is not only white balance. It is also a channel mixer.

Take a look at the tab "colorfulness" and the "gamut compression" in the "CAT" tab. I try to post an example later.

Click to expand...

Yeah, I did play with them, including adjusting individual RGB channels. The problem is, the blue discs are already solid with the choice of "standard color matrix" (whereas they should be half-transparent), gamut compression only pulls the purple/violet back into blue, perhaps similar to what Erik did in LR.


left: default gamut compression, middle: gamut compression bring back the blue, right: using prophoto RGB as input profile

I could also choose standard color matrix as input profile, and set Rec2020 as both working and clipping profile, that would prevent pushing blue into purple/violet, but those discs/circles are already saturated/clipped into solid color to begin with.

I'm with Erik.

Reducing 'exposure' in Capture One shows that you can recover some of the 'true' colours but as they have been over-exposed in thRAW file this may not reproduce the real colours.

Putting the DSCF3187.RAF into RawDigger it shows both Red and Green channels are at maximum and the Blue channel is not far behind.

As displayed in Capture One (the lower image is unadjusted ).


Film Curve = Auto (Pro Neg Std.) and Exposure -2.2


Film Curve = Film Standard and Exposure -2.0




Film Curve = Linear and Exposure -1.1

RG

RetiringGuy said:

I'm with Erik.

Reducing 'exposure' in Capture One shows that you can recover some of the 'true' colours but as they have been over-exposed in thRAW file this may not reproduce the real colours.

Putting the DSCF3187.RAF into RawDigger it shows both Red and Green channels are at maximum and the Blue channel is not far behind.

As displayed in Capture One (the lower image is unadjusted ).

Click to expand...

Thanks. As they're light sources, of course clipping, especially in the center of the circles/discs, are expected. I'm not interested in retaining all the colors, just trying to see how I could best process these to a pleasant output (at proper brightness while still not clipping too much of the colors, or pushing them into a different colors). It's probably a limitation that darktable is still trying to solve.

It's also interesting to see how Fuji cameras handle them very nicely.

sluggy_warrior said:

Erik Baumgartner said:

I think this is just a clipping, not a gamut problem. here is just pulling back the blue channel back (mostly) from beyond the clipping threshold (Lightroom). Same profile and color settings.

Click to expand...

Thanks. It could just be darktable, being a FOSS tool, doesn't have the resources to profile most cameras, thus, its standard matrix is just not as good as the commercial tools.

In your screenshots, although most of the blue are pulled back, the bright green circles now look a little off (greyish)?

Click to expand...

This was just a quickie global highlight/color recovery - definitely over-recovered in places. If i was going to make a proper edit of this RAW file, I would approach this differently with both global and local adjustments in the mix to recover as much color fidelity as possible while maintaining maximum brightness.

His Dudeness said:

sluggy_warrior said:

Here is a typical output in darktable using standard color matrix as color input profile.

Click to expand...

quick tip: you may try the color calibration module to pull colors into gamut.

Will try to show something later

Click to expand...

Here we go (all with standard color matrix)


darktable with filmic and color calibration


darktable with filmic, color calibration and Astia LUT


darktable with filmic, color calibration and highlights pulled back

His Dudeness said:

Here we go (all with standard color matrix)


darktable with filmic and color calibration

Click to expand...

Yeah, sadly, those blue-turning-purple solid circles/discs just confirmed that Aurelien still has works to do The screenshots of C1 by RetiringGuy look much better with correct colors and tranparency.

I'll try to share these on pixls.us to see what darktable tricks I missed. Granted, I don't understand any of the color theories, hopefully will learn something useful.

Erik Baumgartner said:

This was just a quickie global highlight/color recovery - definitely over-recovered in places. If i was going to make a proper edit of this RAW file, I would approach this differently with both global and local adjustments in the mix to recover as much color fidelity as possible while maintaining maximum brightness.

Click to expand...

Agreed, I'd bump up the dark background a bit to provide some context, and reduce the contrast for a slightly softer look, too.

Thanks all for the samples from LR and C1, I looked at them and SOOC, and decided that I'm happy with the soft blue from using the larger prophoto color space. The colors are slightly muted comparing to Astia, although I can crank up saturation, I like the soft look and smooth gradient of the vertical trails near the top. I guess when I have to compress the gamut into sRGB, accuracy is the least of my concern

At least this is what I like for now.


linear prophoto RGB as input profile, bumped up the background a bit with tone equalizer, and filmic for curve

sluggy_warrior said:

I'm processing my backlog and ran into the rare cases of light sources with colors out of the usual sRGB space.

Click to expand...

Without measuring actual colours it's impossible to say are they inside or outside some gamut.

But if they are clipped in raw, they are distorted anyway, and fidelity is out of question.

sluggy_warrior said:

darktable doesn't have an enhanced color matrix for the X-T20, and the standard color matrix results in color clipping badly. I had to switch to a larger linear Rec2020 RGB or linear prophoto RGB but of course the colors then are slightly off.

I fed the RAF file back to the X-T20 and generated the JPGs for different film sims. From the result, looks like Astia is the most accurate, still has plenty of saturation and contrast while staying true to the original colors. Velvia pushed the saturation a little too far, changing saturated blue into purple/violet.

I'm just curious how different RAW processors (LR, C1, ...) handle the wide gamut. Here are the two RAF files if anyone can give me a hand with your favorite tools. Thanks.

https://drive.google.com/drive/folders/1UH8tlVtrJUUR1f_vlSbMkCLWrzCCjS1x?usp=sharing

The SOOC JPGs generated by the X-T20:


Atsia SOOC


Velvia SOOC, bright blue discs turned into purple/violet

Click to expand...

--