Luminosity mask applied backwards by GIMP

I am using GIMP 2.10.12 on Windows 10. I worked through the Pat David article Luminosity Masks at in January. Because I wanted to use the method on a recent photo, I looked at it again and worked though the creation of the masks. However, it seemed that GIMP was applying the wrong channel to a layer.

I followed the article to create a Lights channel and a Darks channel. To examine each mask, I applied them to copies of the base image and used Show Layer Mask. The resulting image was backwards to what I got before and backwards compared to the result in the article: the dark mask appeared like the light mask and vice versa.

The photo used in this post and the article is Mountains as Far as the Eye Can See (cc-by) by the West Arctic National Parklands.

My layers:
An image of the Lights mask from my work:

From the article, “what the three different channels of Darks look like:”

The article images were produced by a much earlier version of GIMP, which is the reason for the difference in appearance.

An image of the Darks mask from my work:

The Lights masks from the article:

Perhaps I am not using the right way to display the masks, but I have checked multiple times that I am applying the correct channel for my layers, and I always get this result. I have also reinstalled GIMP twice.

The tutorial works, and Gimp also works.

What the tutorial shows is the masks as selections mask/channels. If you use them a layer masks, you are making a composite image with the image below, and this isn’t really the mask.

The easiest way to check the mask is to drag it from the Channels list to the canvas. This creates a new layer which is an exact copy of the mask (delete the layer afterwards).

1 Like

Thank you very much.

Seems like something else is wrong, then. To illustrate the use, the article shows the adding of the DD layer mask and Colorizing with teal:

Here is my attempt to do that:

Notice the difference is the icon of the layer masks. You can see the icon of the DD copy layer right above the masked layer. Compare those to the article’s mask icon. Notice also that the teal colorizing is applied to pretty much the whole image, not just the darker tones.

Hi @Underexposed,

Make sure to turn off the “Invert mask” option when adding the mask.:


1 Like

Additional remark: one of my statements above isn’t 100% true, the tutorial doesn’t work in “linear light”:

  • The "Darks’ mask is obtained by subtracting the “Lights” from “Select all”. This gives a very different result in linear light.
  • In perceptual mode you can obtain the same result as the tutorial’s method with just Color>invert and in Linear mode this can be replaced by Color>Value invert.
  • Another equivalent is to use Curves in perceptual mode and use a top-left➜bottom right diagonal. An advantage of the Curves is that you can save the luminosity masks as named presets that you apply to the “L” mask.
1 Like

It’s the little things that kill you. Thank you for pointing that out.

I did see the difference in using linear light, and I see how invert works. Your third item, does that mean to apply to the “L” mask to get “D”?

The creation of the M masks puzzles me. In the Tony Kuyper article, he says to select all and then, for M, subtract L and D. That does not, for me, produce the mask in Pat’s article. I get a solid black mask.

However, I have never been able to get the MM and MMM masks using the method of LL mask, Channel to Selection, and DD mask, Intersect with Selection. I get solid black masks. If I use the Kuyper method, select all, subtract, e.g. LL, subtract, e.g., DD I get the MM mask. Similarly, with the MMM mask.

Sorry, Boris. I guess I don’t pay close enough attention. Thanks for the tip on the invert.

This helped me when I got confused. Actually, the whole thread is quite informative.

I agree with you, the M obtained by subtracting L and D is all black. Remember;

D=1-L (select all and subtract L)



However, in Pat’s article this is not how you obtain M. M is L interstected with D (or vice-versa) and this gives the expected result. You’ll note that the M obtained that way tops at 50%, but you can get a 100% selection range:

  • L to selection
  • D intersect to selection (gives Pat’s M)
  • Save to channel
  • Add saved channel to selection (so selection that was M is now 2*M)
  • Save to channel as M, and erase previous channel.

Yes, that explains it clearly.
Question: when you want to link to a specific spot in a conversation, like you did, how do you create the link?

You have created another way to make MM and MMM that requires fewer steps. I think subtracting L and D works in Photoshop, which is what the Kuyper article is about, because of differences in how it works.

I use a Pat David script, “re-coded by Saul Goode,” and it works well in GIMP 2.10.12.

1 Like

If the link is on a separate line or paragraph, the forum should expand it into a preview box. It doesn’t always work for external links.