Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Making a Selection from a Channel
#1
Hi there, thanks for reading this!

I have an image (layer) in b&w  I want to use as a mask. In Photoshop, I used to make sure the layer is active and just press "strg+alt+2" or make a selection from the RBG or Green channel in one click. This gave me a selection that is hard in the bright white areas and softer in the areas where the light is less bright. I used this simple method to edit the highlights in my images with absolute precision, since I can paint on the b&w layer and exclude certain areas as well. It was so quick an easy and I was wondering if there is fast way to get the same selection out of gimp?

Thank you in advance
Reply
#2
(03-13-2025, 11:23 AM)cloudescaper Wrote: Hi there, thanks for reading this!

I have an image (layer) in b&w  I want to use as a mask. In Photoshop, I used to make sure the layer is active and just press "strg+alt+2" or make a selection from the RBG or Green channel in one click. This gave me a selection that is hard in the bright white areas and softer in the areas where the light is less bright. I used this simple method to edit the highlights in my images with absolute precision, since I can paint on the b&w layer and exclude certain areas as well. It was so quick an easy and I was wondering if there is fast way to get the same selection out of gimp?

Thank you in advance

This is a "luminosity mask". The most basic way is to open the Channels list, right click one of the R/G/B channels, and "Channel to selection". You can get more fancy, making a grayscale version (Colors > Desaturate > Desaturate or Colors > Components > Extract components and pick a relevant component (LCH luminosity, for instance)), taking any of the RGB channels, and then applying Curves to select a specific range. There are even scripts for this, for instance my own ofn-luminosity-masks.
Reply
#3
I also have a question about channels so figured I'd add here.

Context:
I'm learning how to conduct image integrity checks on scientific articles. One way to check for duplications is to copy the 2 suspected duplicate images into a new file, one in the red channel and the other in green. Reducing opacity and overlaying them then makes it clear what matches and what doesn't.

Issue:
I have my two sections in separate layers on the same canvas but cannot figure out how to lock one into the green channel and the other into red. Each time it just changes the entire canvas and all layers to the selected channel. I've tried decomposing and selecting but cannot seem to get that to work either.

Any suggestions?
Reply
#4
(03-20-2025, 02:50 AM)exxo Wrote: I also have a question about channels so figured I'd add here.

Context:
I'm learning how to conduct image integrity checks on scientific articles. One way to check for duplications is to copy the 2 suspected duplicate images into a new file, one in the red channel and the other in green. Reducing opacity and overlaying them then makes it clear what matches and what doesn't.

Issue:
I have my two sections in separate layers on the same canvas but cannot figure out how to lock one into the green channel and the other into red. Each time it just changes the entire canvas and all layers to the selected channel. I've tried decomposing and selecting but cannot seem to get that to work either.

Any suggestions?

You can do like everyone does, set the top image to "difference" mode. If the images are identical you get black, otherwise you get something more visible.

If you still want to use color channels:
  • You can replace each image by one of its channels using Colors > Component > Extract component
  • Better solution:
    • Put each image in a group
    • A layer in the group above each image
    • Bucket fill that layer with the channel you want to extract (red, green, blue...)
    • Set the color layer to Multiply
    The perk of this more complex version is that you can just toggle the visibility of the color layers to see the complete original image
Reply


Forum Jump: