Blue color vs. over exposure

Software RawTherapee
,
1

I am sometimes experiencing slightly odd results when my Nikon D600 runs into very saturated blue light that approaches over exposure. An area being hit by light may in fact end up becoming darker than areas not hit by this light. I believe it’s a result of RawTherapee’s way of dealing with color saturation (increasing saturation causes colors to always become darker in RawTherapee when using the “Saturation” slider, and maybe the same algorithm is used inside the color profile section?)

Here’s a piece of white paper on top of a blue LED. For each picture I raise the exposure level, and then odd things start to happen:

www.robotplanet.dk/f/rawth/blue/std00.jpg

www.robotplanet.dk/f/rawth/blue/std01.jpg

www.robotplanet.dk/f/rawth/blue/std02.jpg

www.robotplanet.dk/f/rawth/blue/std03.jpg

www.robotplanet.dk/f/rawth/blue/std04.jpg

Now, disabling the color profile section makes things a bit more normal:

www.robotplanet.dk/f/rawth/blue/no00.jpg

www.robotplanet.dk/f/rawth/blue/no01.jpg

www.robotplanet.dk/f/rawth/blue/no02.jpg

www.robotplanet.dk/f/rawth/blue/no03.jpg

www.robotplanet.dk/f/rawth/blue/no04.jpg

So try disabling the color profile if you have this problem. But watch out: Disabling the color profile may also result in odd colors, so this is not a perfect work-around.

2

A friend of mine has observed the same with a Sony ILCE-5000 camera in RawTherapee.

If anyone knows a different work-around than my suggestion, please do tell.

3

Did you try on Darktable ? It could be a wrong color profile.

4

I’m on Windows, and sadly it seems there’s no Windows build available according to their own website. But this behavior has been confirmed with a Sony camera too (by a friend of mine in switzerland), so it might be more than just my setup.

Your suggestion to try other kinds of software is good though, so I’ll see what I can find, and give that a shot. (In fact, I just tried Adobe Camera Raw 4.6 which resembles RawTherapee with color profile disabled: The spot of light does not become darker than the surroundings.)

PS: Here’s the raw file in case anyone wants to toy around with it: www.robotplanet.dk/f/rawth/blue/blues_leds_under_paper2.dng

5

I’ve this problem too with my Nikon D90. It helps by using a DCP profile where you can check “Look table” under “Color Management”.

6

Advising the use of “no profile” is bad advice.

We can take a look if you upload the sample raw + PP3.

7

Here’s one of my outtakes.
DSC_0407.NEF.pp3 (11,2 KB)
DSC_0407.NEF (18,6 MB)
Crop 100%
grafik

8

Here is my raw file and settings based on the “Bundled/Default” preset that comes with RawTherapee, but with contrast and brightness set to “0”, and with exposure at +4 where the phenomenon is visible.

http://www.robotplanet.dk/f/rawth/blue/blues_leds_under_paper2.dng
http://www.robotplanet.dk/f/rawth/blue/blues_leds_under_paper2.pp3

9

You can use Darktable with Windows. Read this post: darktable for Windows

10

@TooWaBoo

Perhaps it can help
Nikon D90.txt (1.5 KB)
Rename it *.dcp

11

@gaaned92, Thanks.
I’m using the Adobe DCP profile “Nikon D90 Camera Neutral.dcp” and check “Look table”. This gives a similar result.

12

Interesting read:
https://www.ludd.ltu.se/~torger/dcamprof.html#make_profile_deep_blue

2 Likes
13

I just installed darktable and used the default settings, but boosted the exposure and contrast a bit. I immediately got results that are far more odd than RawTherapee:

1140×940 143 KB

So far, the most realistic output has been my original (bad) suggestion: Disabling the color profile, which is a hazardous approach, but probably what I’ll be doing whenever I’ve been encountering LED lights in the wild.

14

I think this is because blue channel contributes far more than red and green channels to the luminance during converting the picture to a colorspace e.g. sRGB (Y ~ 0.21R + 0.72G + 0.07B). Investigating the RGB values in the different parts of your image supports this hypothesis as it shows that the values of red and green channels are zero in the deep blue area while their values are not zero in other areas.

You can read more in the following Wikipedia articles:

15

That is an old and known problem with many cameras, saturated blue light and the color profiles shipped by default: They are resulting in broken colors. In darktable try opening the input color profile module and setting the gamut clipping setting to Rec2020.

16

The default working profile is ProPhoto. Your PP3 used sRGB as the working profile. Changing it back to ProPhoto, or to Rec2020, solved the issue:

i
i1920×1023 618 KB

2 Likes
17

I’ll stick to your suggestion for now, and meditate on the other settings and the documentation for now. It seems right.

Actually I just realized that if I pick Bundled Profiles / Default (which uses “auto-matched camera profile”, working profile ProPhoto and output profile RT_sRGB) and then disables the “Base table” checkbox, I get really nice results. So maybe the “Base table” thingie is what I don’t like. (Will read up on this now… )

The reason I’m experimenting this much with my Nikon D600 images is, that I’ve never reached a point where I get the same fantastic color reproduction as the Fuji S2/S3/S5 Pro series that I used to work with. These cameras even gave perfect colors when using out-of-camera jpeg, which I find pretty surprising considering what camera jpeg files usually look like for most other cameras.

I have a gut feeling that Nikon (using a Sony sensor btw) are sacrificing color precision a bit to get lower noise levels at high ISO values, which is where D600 beats the Fuji S5 any day. On a sunny day I’d prefer the S5 Pro, but at night the D600 wins.

I’m just happy that RawTherapee lets me tweak so many settings. It’s quite educational.