A study in blue

Two years ago I shot a photo at a concert that I have since then used to test various image editing software. It has everything you don’t want: A lot of blue LED light (but also other lights), clipped areas, and dark shadows that easily lose contrast.

RPN_Kick-In_20180830_3392.NEF (9.6 MB)

At the time I shot it I was using Lightroom and it annoys me to no end that it just dealt with it and gave me a usable image and I have no idea how.

(Lightroom 6)

In darktable I managed to get this edit that I like, but I had to use a few very ugly hacks to get it like this. such as

  • Filmicrgb with v3 color science
  • Channel mixer in the pipe before input profile
  • Rec709 as input profile (which makes no sense), the default matrix for my Nikon D300 gives all kinds of nasty artifacts.

(darktable 3.2.1)
RPN_Kick-In_20180830_3392_01.NEF.xmp (12.5 KB)

(darktable 3.2.1, simple edit with scene-referred workflow for comparison)
RPN_Kick-In_20180830_3392_03.NEF.xmp (7.6 KB)

I would of course prefer to make this look good without weird tricks, so I’m hoping that you have suggestions, tricks or other ideas on how to handle an image like this gracefully and get the most out of it.

If I remember correctly RawTherapee did reasonable as well, mainly because I could use Adobe’s DCP profiles with RT.

All these files are licensed Creative Commons, By-Attribution, Non-Commercial, Share-Alike 4.0.

6 Likes

In rawproc:

I have a SSF LUT profile for the Nikon D300s, which improves the blue handling a good bit over the default matrix. Took an aggressive lift with the filmic curve to bring the image up to pleasing viewing, then a bit of nlmeans denoise prior to resize/sharpen for output.

3 Likes

That’s pretty good! I forgot about your LUTs, my camera should have the same sensor as the D300s (and probably the D90) so those SSF’s may be valid for that as well. Can I download those somewhere?

Right here:

I’m collecting all the SSF data I can find and posting it here if the license permits.

2 Likes

Here is a one-click solution using the new color calibration module in dt 3.3.0-git+2036

Yes, of course I could have worked with it more — but wasn’t it interesting
what just one click would do?

Have fun!
Claes in Lund, Sweden

6 Likes

Here a “manual” several clicks solution in darktable.


RPN_Kick-In_20180830_3392.NEF.xmp (7.3 KB)

2 Likes


RPN_Kick-In_20180830_3392.jpg.out.pp3 (13.9 KB) RawTherapee 5.8

Thanks for this challenge!

1 Like

RPN_Kick-In_20180830_3392 RPN_Kick-In_20180830_3392.NEF.xmp (23.8 KB)

Filmic v4, color calibration, and 2 years of color science R&D:

RPN_Kick-In_20180830_3392.NEF.xmp (22,7 Ko)

10 Likes

I tried to go for that Bon Jovi - It’s my life look with DT Master :slight_smile:

1 Like

Here is my take with darktable master (color calibration to compress gamut, filmic with highlight reconstruction that blooms)
i tried to preserve as much as possible the blue mood and to get smooth transition between shades of blue

Another attempt, using the “colorfulness” tab of color calibration with a negative value on blue instead of gamut compression

4 Likes

Yes! That is a very valid point [and perhaps worthy of a thread of its own].
Do we really strive for showing clinically “correct” white points & balances & colours
— or do we want to preserve a mood?

Just imagine the scene where Tramp invites Lady to a candle-lit dinner at Tony’s…

Have fun!
Claes in Lund, Sweden

1 Like

Cool to see what everyone makes of this, thanks!

@KristijanZic @rawfiner can you maybe upload the xmp files for your edits as well? Curious to see how you managed it, I especially like the last one.

ART

second trial using @ggbutcher profile

1 Like

@Egocentrix the xmp is included in the jpg (you can import the jpg file in darktable the same way as you would do with an xmp file) :slight_smile:
After some other tests with similar pictures, I think the most efficient way for me to handle those blues is to compress gamut and use a negative value of blue in colorfulness at the same time: using more the gamut compression if I want the pixels to be darker, or more the colorfulness if I want them to be brighter. Though, using colorfulness can give hue shifts, which may or may not be problematic depending on the image. I also usually lower the default middle-tone saturation in filmic for such images.

RPN_Kick-In_20180830_3392 RPN_Kick-In_20180830_3392.NEF.xmp (12.0 KB)

Huh, didn’t know that. Neat. I’ll try some dt3.3/color calibration edits tonight.

Halos look much better with “clip negative RGB from gamut” = off.

For the human perception, it is an advantage that the same things always have the same color. Therefore, the human eye (or better brain) adapts to different lighting conditions trying to identify the “real” colors (that do not exit as such). When adapting the whitepoint, we try to imitate this behavior, to prevent objects from looking “strange”. But the human adaptation has limits, objects at sunset still look somewhat reddish and concerts with lots of blue led lights make the surroundings still appear bluish.
So, I think it is more than preserving “mood”, it’s about making a scene look right according to the human perception.

2 Likes

Interesting discussion. I mostly like to work with captures in their original color context, as the best colorimetric effort of the camera. However, with some images I then like to “go wild” and see what some abstraction will render. Monochrome is a frequent departure, but I have a few early images where I just put them in GIMP and threw G’MIC operators at them until they looked Quite Different… :laughing: