As shown below, an extracted channel (here blue) gives a brighter result when using the "Channel to Selection" from the context menu in the Channels panel and creating a mask from that selection compared to using Colors>Components>Decompose or Color>Components>Extract Component (which both give the correct result):


Colors > Components > Extract Component > RGB Blue:


Channel to Selection:

Is this intended and what is the reason?


I assume that after "Channel to selection" you bucket-filled the selection with I white on a black layer.

On Gimp 2.8 the behavior is as you expect(*). 2.8 works all over with 8-bit gamma-corrected values (so that there are more values to describe the dark tones).

In 2.10 things are different. Internally all computation is done in floating-point, using non-gamma-corrected values. When you copy the channel to a selection the plain channel data is copied, and since the selection is not really an image, there is no reason to gamma-correct it.

There is still a way to apply this gamma-correction, though. Assuming you have a selection from the Blue channel:

  1. Create a layer filled with black
  2. Add a new layer, and bucket fill the selection with white (you get the "too-light" result)
  3. Change the blend mode of the white layer from Normal to Normal (legacy) and you get your expected result (basically the "legacy" blend modes skip the gamma-correction).

(*) "Correct" is a bit strong here. There are plenty of ways to create monochrome images, all as a valid as the others.

  • Thanks. I did not create a white or black layer, I simply did Add Layer Mask and checked "Selection" but I'm sure it's the same. – ChameleonScales Jul 13 '20 at 8:59
  • Are you sure its gamma corrected? And not inverse gamma corrected or profile corrected? Also its correctish assuming you manipulate images that are created with realworld physics and viewed with human eyes; Yes... But you might not – joojaa Jul 13 '20 at 17:11
  • I did not apply a gamma correction, at least not knowingly. Is there a setting I may not be aware of? – ChameleonScales Jul 17 '20 at 12:43
  • No that's just the way things work in 2.10, and why they kept the "legacy" blend modes for when you depend on the historical behavior (which itself is wrong in many occasions, they didn't change this without good reasons). – xenoid Jul 17 '20 at 12:50
  • Thanks. Ok, I made a stupid decision to duplicate post. I invite you to read this forum post and tell me if that's ok for you if I delete the current StackExchange post. – ChameleonScales Jul 17 '20 at 14:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.