Strange color artifacts on red, blue or violet

I have noticed that rawtherapee produces stange artifacts on red, blue, or violets colors, for certain hues. For a better understanding I put in the following an example:

In this example, it appears blue pixels within the violet area. Actually, this area is only violet. I can have the same issue within red or blue areas
Do you have an idea about the root cause of this problem? How should I solve it? I have tried various dcp profile and it did not solve it…

I put here a link to the raw file. Please, do not publish the face of the kid on the website.
NEF_FILE
Thank you in advance for your answers!
olivier

Salut, Olivier!

I am not really certain what is your complaint :slight_smile:

  • Could it be the longitudal, lighter parts of the cat-like animal?
    I believe that those simply might be a fold, catching more light.
  • Or, could it be the heavy colour noise?
    If so, that is because you shot the image @ ISO3200 and you have to work more with noise reduction.

Have fun!
Claes in Lund, Sweden

Well I tend to think it is just noise but… did you actually mean that some blue parts should be purple? I do not see any purple (violet) areas, only several shades of blue.

To me it looks like color noise but it seems to be isolated only on the animals body which does seem odd. The outside blues just have typical luma noise. I am not at a computer currently able to test as I am at work but maybe defringing or more aggressive chroma denoise can fix?

This does not look like any artifacting I have ever seen.

As @Claes, @betazoid, and @blj said, it’s the noise combined with very high saturation. If you desaturate the color in question and remove the noise, it will look good again:

To me it looks like the noise is pushing colours out of gamut, or into clipping when the colour profile is applied.

Try increasing ‘false colour suppression steps’ in the raw tab.

Difficult to assess the problem without your pp3 profile. Surely you increased the saturation


DSC_1065.jpg.out.pp3 (12.7 KB)

Thanks for your feedbacks. It appears that changing the tonal curve from film to perseptual helps to alot to remove these artefatcs.

Let’s look at it in stages, each one applied after the previous one:

  1. image loaded and neutral profile applied (Amaze demosaicing algorithm):


    With Auto-matched camera profile + Base table

  2. noise reduction applied, default settings, Lab color space:

  3. noise reduction, RGB color space:

  4. noise reduction, RGB, aggressive mode:

  5. RCD + VNG4 demosaic:

  6. RCD + VNG4 demosaic and False color supression (worse result):

  7. defringe, default settings:

It seems it’s a strange or unusual problem, but I don’t think it’s a bug. You just have to know what it is and search a way to remove it.

EDIT: sorry, I forgot to say that in the last step there is NO false color supression applied

1 Like

Having recently investigated such in other images, I’d agree with @Iain that these are out-of-gamut colors being squashed in a color transform.

I’ve tried to do things like reduce color saturation, and even modify the matrix transform in the camera profile, but if you can’t locally apply such they tend to adversely affect the color in the rest of the image.