No graduate saturation (example pic included)


Perhaps this will help.


That’s cool. The actual thing that is happening with the picture.

I am playing around with that now and get a better (less dark blue, more gradient) when unchecking ‘Look table.’

I have asked the person in the photo if it would be okay to upload the photo here to learn about colour profiles, and it is accepted! So here we go. :grinning:

DSC_4099.NEF (24.4 MB)
This file is licensed Creative Commons, By-Attribution, NonCommercial, Share-Alike
(Creative Commons, By-Attribution, NonCommercial, Share-Alike)

My best result while using the Nikon software profile:

Face looks a bit too green. Blue gradient is not at it’s best either.
DSC_4099_RTc9.png.out.pp3 (12.2 KB)
Nkx_D7500_6022567_719_1_0_0_04_12_00_00_01_06_12_12_00_0320_0_7_2_65_476_00ff00ff010002_0000ffff.icm (1.6 MB)

My best result while using the Adobe software profile:

Face is way too red. Blue gradient is restored although it is a bit light.
DSC_4099_RTa5.png.out.pp3 (12.3 KB)
Used “Nikon D7500 Camera Flat.dcp” from Adobe pack here.

I am playing with “Nikon D7500 Camera Standard.dcp” and “Nikon D7500 Adobe Standard.dcp” now. Since I just discovered those DCP-options @TooWaBoo mentioned, I am seeing where that brings me, but I always prefer a systematic approach that is ‘always’ working. Well let’s say a method that I can apply to all dusk/night shots at least. Or is that just a bridge to far to wish for? Should I for each image try about 3 profiles and switch around with the 4 DCP-settings?

(Morgan Hardwood) #23

Using the NIKON_D7500_neutral.dcp I made, almost default settings except for resizing (there is no need to upload 95MB PNGs), setting highlight reconstruction to “Luminance Recovery” and highlight compression to 50.
DSC_4099.jpg.out.pp3 (10.8 KB)

And you’re free to adjust the lights to taste:



Thinking back to the link @ggbutcher posted about the blue…
When playing with the ‘Blue Primary’ in Adobe DNG Profile Editor, Color Matrices tab…when moving it towards the left, cyan, gradient comes back. In a lesser extend also when moving it to the right, purple. And also when decreasing the saturation. So I opened “Camera Neutral (Nikon D7500)” and decreased saturation with 20.


Not bad. It must be the night scenery that requires a “night profile” really where the range of blue is decreased to bring back the gradient. Don’t know if I used the right method of color remapping though.

(Ingo Weyrich) #25

That photo would be also great to demonstrate the latests sharpening and demosaic improvements in RT.

Do you allow to use the image for showing the improvements?



Hi, what would that mean exactly?


@SaturnusDJ He is asking you if he has permission to use your image to demonstrate the latests sharpening and demosaic improvements in RT. If it is okay, add a license to it; e.g., see.


Thanks for the info.

Applying CC in original post.

(Glenn Butcher) #29

Been thinking a good bit about this one, as I’ve run into the same treatment of blue in shots I’ve taken in auditoriums. I guess cobalt blue is a popular lighting color… Anyway, I read the dcamprof documentation about this and decided to try the fix, manually adjust my camera profile blue Y in the matrix to -0.1. Here’s a link to the dcamprof documentation, a read I highly recommend:

I started with the calibrated camera profile I made using Elle’s instructions. Here’s it’s xyY gamut plot:

Yep, bluest-blue is way off the visible, which makes for a long gamut transformation.

First step was to convert the profile to dcamprof’s json format:

$ dcamprof icc2json Nikon_D7000_Sunlight.icc Nikon_D7000_Sunlight.json

which gave me this:

“Description”: “Nikon_D7000_Sunlight.icc”,
“Copyright”: “No copyright, use freely.”,
“UniqueCameraModel”: “D7000”,
“CalibrationIlluminant1”: “D50”,
“Whitepoint”: [ 0.780472, 0.809677, 0.634842 ],
“ForwardMatrix”: [
[ 0.756226, 0.222610, -0.014633 ],
[ 0.277954, 1.010818, -0.288773 ],
[ 0.017166, -0.208313, 1.016052 ]
“ProfileConnectionSpace”: “XYZ”,
“RedTRC”: 1.000000,
“GreenTRC”: 1.000000,
“BlueTRC”: 1.000000

In the “handling extreme colors” section, Anders describes the situation, and points to the Y component of the blue primary as the starting place to address it. dcamprof make-profile has a -y switch that allows you to set the minimum Y for all three channels; me, I figured I could just change the json and turn it back into a ICC. So, I hand-edited the -0.288773 to -0.1 based on the guidance he gives in the “Deep blue handling” section. I figured if that were too much, the process was cheap to iterate. The command to make the round trip back was:

dcamprof json2icc Nikon_D7000_ModBlue.json Nikon_D7000_ModBlue.icc

I moved the ModBlue ICC file to my “profile zoo” and opened a test image using the original camera profile:

(Image copyright 2018 Glenn Butcher; all rights reserved)

Yuck, that spotlight blue just gloms into a single amped color. So, I then assigned the modblue camera profile instead of the original one, and got this:

(Image copyright 2018 Glenn Butcher; all rights reserved)

Much better. Here’s the gamut comparison:

Yes, my gamut plot tool is hobbled in height… But you get the idea, now, the bluest-blue is a lot closer to the visible range and less of a struggle for the relative_colorimetric rendering intent.

There’s probably more to do here that involves a LUT, but I thought this was pretty good by itself. dcamprof is pretty handy in this regard, handles both icc and dcp profiles.

FFT (Food For Thought)…

Adobe Profiles and the Color Management Chain

The blue is essentially a camera metamerism. If you check out ACEScentral, you will see that it is common for certain sharp band lights to trigger the phenomena.

The result is exacerbated when using a wider gamut reference, especially given the fact that camera primaries are cheated math variants, which yields artificial values when attempting to locate the actual chromaticity of the original source. All of this compounds with an unfortunate display referred pipeline as the intensity is likely also being incorrectly displayed.

TL;DR: Think camera metamerism and it makes sense.


@troy_s From your perspective, what is a good way to deal with this specific problem?


I guess I would start with understanding that the captured value is a non-colour | non-data. That is, just like an RGB triplet missing a component or two, the value ceases to be data (despite all of the contortions silly folk go into regarding “highlight recovery”).

So I suppose the question is how to salvage broken values. The current ACES solution is to change the blue primary as an intermediate step. It seems to work well enough given the edge case nature of the phenomena. More discrete view transforms would be more improved.

If one were in a scene referred pipeline, in theory, if the relative exposure intensity is large, the desaturation component of a decent viewing transform might handle it well enough. On a case by case basis, the most alien looking metamerisms will occur when the bulk of the image has a certain saturation, and the questionable values spike. A secondary grade could pull those sorts of instances in.

Given it is a busted up bit of non-data, there isn’t a terrific solution, as the areas around the data are also likely mangled up. It essentially drifts into visual effects / post production repair work in those instances.


I’ve exported two images from rawtherapee , no profile and camera standard.
Than I’ve played a little with gimp

So the “no profile” was usefull for the lightness(LCH) blending mode and the “camera standard” for the hue(hsv), it looks very similar to the jpeg version :slight_smile:


Would like to say thanks to everybody who contributed to the topic. Some stuff is pretty advanced but I’ll get there along the photography way.


Except the JPEG isn’t a ground truth; the JPEG simply has a similar broken rendering to what you are doing, which is equally broken.

It’s really just broken non-data in those regions.


Yeah. Correct me if I am wrong, but what I tend to see is the deep blue being converted to a more purple colour, as indeed can be seen in the JPG. And not only here but in other shots also. I just don’t remember the lights being purple in real life.