ART - bug combining masks

[WestCountry]

I discovered this bug with combining masks in ART 1.25.6 last night, and have checked it’s still in ART 1.25.8 this morning.

As an example, here is an image where I want to darken the sky.

I then select the sky with a colour similarity mask, but this does not select all the sky (and selects some of the train, but that’s not important for the bug).

I then add a brush mask on top (in Add mode), to cover the remaining bits of the sky.

If I deselect the Colour Similarity mask (to just show the brush mask), the brush mask is shown filling the entire image.

To get around this, I have to change the brush mask from Add mode to Intersect mode, to see just the brush mask.

As an aside, at this point my cursor got stuck as the brush tool. The only way I could stop this was to open a different image, then re-open this one.

Then I set Highlights -100, with the result below:

As you can see, the edges of the brush mask are showing up, the sky is darkened much more where the brush mask is, than where the colour similarity mask is.

I then turned Posterisation up to 6, to set the mask to either off or no. However, this shows the colour similarity mask at approx 50% opacity, but brush mask 100% (even where not overlapping colour similarity mask) - even though both masks are set to 100% opacity in the tool panel.

Please could this bug be investigated, and fixed if possible, so adding on a brush mask doesn’t reduce the opacity of the colour similarity mask?

Here is the RAW file, and the .arp sidecar file, so the bug can be verified.

2025-05-03-15h46m12s-4501.NEF (26.4 MB)
2025-05-03-15h46m12s-4501.NEF.arp (16.1 KB)

As well, last night I tried adding an Area mask to the colour similarity mask, to see if that would work where the brush mask didn’t. This behaved very strangely, if I set the Area mask to Add mode it acted as if it was in Intersect mode

Whereas here is the colour similarity mask on its own

And here is the Area mask on its own

And if I turn up Posterisation to 6 it shows only approx 50% opacity in the intersected areas – despite being in adding mode

These Area mask settings are included in the above .arp file and RAW file, so they can also be verified.

[Thierry007]

Hello, I don’t think there is a bug, it’s just that the way you are using the masks is not correct. I am attaching my processing with the .arp file.

2025-05-03-15h46m12s-4501.NEF.arp (15.6 KB)

[WestCountry]

Hello,

Thank you for your reply, and for testing out the masking on the RAW file.

I have checked with the ART website though, and I am using masks the correct way – according to this section in Local Editing with Layers and Masks | ART raw image processor

The Add mode works correctly when using a brush mask to erase part of a mask, which is what is in your .arp file – adding this Colour similarity mask

to this Brush mask (using the eraser)

to give this combined mask

Which works correctly – although as it does not cover some of the steam in the sky, this is still too bright.

However, if I then paint the mask over the gaps in the sky, using the brush tool (not in eraser mode), to give this combined mask

then when I stop showing the mask, then the brush edges can be seen in the sky

This is the problem/bug I have found with the Add mode of the brush mask – it works fine where you erase parts of other masks, but does not work correctly where you try and paint in missing parts of other masks.

My processing profile to verify this
Mask brushed over gaps in the sky.arp (18.1 KB)

[Thierry007]

Here, I just added a second color similarity mask to the smoke.

[agriggio (Alberto)]

Hi,
Looks like a bug in the colour similarity mask, which doesn’t reach 100% opacity in some cases even when it should. Everything else you report seems to be normal from a quick glance. I’ll investigate further though and report back when I have more information.
Thanks for the report!

[WestCountry]

Hello Alberto,

Many thanks for your reply, and for investigating this.

Apologies if my report is a bit long-winded, I didn’t want to leave anything out. Although I did miss out that I’m using ART 1.25.8 on Linux Mint

You are correct that the bug seems to be in the colour similarity mask, as if I create one and then set Posterisation to 6, the mask shows approx 50% opacity - even without adding a brush mask to it.

[agriggio (Alberto)]

So, there was a rounding error in the application of posterization, that was mostly visible with the colour similarity mask. It should be better now, but this doesn’t mean that you won’t get visible brush strokes when retouching a colour similarity mask with a brush mask. This might be perfectly normal. You can fix that by using the mask postprocessing curve, that’s exactly one of its use cases.

HTH

[lphilpot (Len Philpot)]

@agriggio

this doesn’t mean that you won’t get visible brush strokes when retouching a colour similarity mask with a brush mask. This might be perfectly normal. You can fix that by using the mask postprocessing curve, that’s exactly one of its use cases.

Can you point me to some instructions (even high level) on how that’s done? Apologies for my denseness…

[priort (Todd Prior)]

It would be my understanding that you are just adjusting the contrast of the mask…so playing with that would blend things in or maybe exaggerate it…its easy to see…add an area mask with a gradient…set a wide feather range and then play with the contrast curve in the post processing…you can see how it controls contrast visually pretty well there…and thus how you could use it…

[agriggio (Alberto)]

Hi,
basically what @priort says. Here’s a quick video demo:

[WestCountry]

Hello Alberto,

Many thanks for looking into this, and correcting the rounding error.

When you say it should be better now, do you mean it will be better when ART 1.25.9 is released, please? Or is there a way to use ART with this sorted out before then, please?

Thanks also for the video of how to use the mask postprocessing curve, that is really helpful and makes it much clearer

[agriggio (Alberto)]

Yes, you can wait for 1.25.9 or build from source…

HTH