How to understand channel mixing in color calibration of darktable.
Is it that in “ouptput R” you only increase or decrease the red values?
And with “input R” only the pixels that already contain red are affected?
With “input G” only the pixels that contain green, while the pixels that do not contain green remain unaffected by this slider?
And the same is true for “input B”, with which (in “output R” of course), only the pixels that contain blue will have their red parts changed?
But since each pixel contains more or less all three colors R, G and B, this channel mixer must consider the amount of the respective color components.
I assume that, for example, with “input G”, the pixels that contain a lot of G will be affected more than those that contain little G.
Do I understand this correctly?
If you go to “R” (output R), you have 1.0 as the defalut. That means you will get 100% of the red channel. If you go to 1.5 you will get 150% auf the original R channel value.
If you increase input “G” from 0.0 to 0.1, you get additional 10% of the green channel value as output in the red channel, i.e. you convert some green to red.
In the way you can even completely swap channels.
I hope that helps.
Check Boris YT channel. Episodes 55 and 56 will go into detail on channel mixing. Actually watch everything from 61 backwards. You will learn a lot about dt and about processing images in general.
Yes, Boris’ videos are great. I have really caught fire. I watched all the chapters concerning the channel mixer. Channel mixer is certainly not the most important module of dt - but I struggled for a long time to understand what was happening.
I am slowly feeling my way forward.
Thank you for this PDF - looks interesting, nice graph - but my mathematical inclination is not strong enough for that. I need something sensually comprehensible.
Is it that you can reduce blue from the skin color without removing too much blue from the whole image by reducing the value for input R in output B?
Or to remove red from blue sky: in ouptut R reduce input B?
I like using the channel mixer for this kind of thing occasionally, but always seem to forget just what does what and have to work it out again
Yes. But, to take the first example, after reducing the input R you often need to increase the input B and/or input G sliders by the same total amount that you decreased the input R. All this is still in the output B tab.
The point is to keep the ‘total’ amount of B in the image the same.
If you tick the ‘normalization’ box it will automatically do this, spread between the input channels that you’re not adjusting.
Hello @priort
this link is a revelation, of course not light fare, but very very descriptive. Anyone willing to understand channel mixer will get everything they need to know here. Great!
Thank you very much for the link.
