A study about using RGB channels in overlay mode to manipulate details in portraits
Inspired by a video of, who else could it be, our very own @patdavid, namely this one:
In the video he uses the blue channel as a check layer for retouching skin, as the skin details seem to be more visible in that channel. Looking around the internet one learns that the red channel seems to have less detail.
So my original idea was to use grain extract mode on a blue channel on top of a picture to get the skin details out of the picture and then using a red channel in grain merge mode to bring back lost contrast.
It works but is way to strong when using 100% opacity on the color channels and you have to play with opacities.
So I switched both color channels to to overlay. Now the blue channel is inverted to get a contrast reduction, especially in the details we want to hide. And the red channel is used as is, again to bring back detail.
To me this is like subtracting the blue channel and then adding a red channel. Thus in the example picture below using the Mairi portrait, I show some possible combinations, with the code how to read the picture in the lower right corner.
E.g. R-G means both the red and green channels on top of the original picture set in overlay mode with the green channel inverted. B-0.5(R+G)* means all the color channels are visible on top of the original picture. all set to overlay, with the red and green channels inverted with an opacity of 50%, and the blue channel with an opactiy of 100%.
As I said this is just a study, which I found quite inspiring. Maybe somebody else will also like it and may find some use for it. I certainly do see some use.
Neat! I’m going to give this more of a try a little later if I can clear up some time.
It’s a classic simple trick from a long time ago to use a copy of the red channel in Overlay over the base image to give the skin tones a nice little ‘boost’. What were you looking to achieve in this case, starting out? Enhancing the skin to more pleasing tones?
I am practicing my skills at postprocessing portraits. Looking around the net it always comes down to using the brush. I am lazy, and so I am always looking for ways to select certain areas automatically. Thus I love mask of all kinds!
More specifically I wanted to have very sharp portraits without also enhancing (through the sharpening processes) the skin impurities to much. I’m not looking for that unreal puppet look, just something along the lines of your Mairi tutorial or your skin retouching with wavelets tutorial.
I have already found a way using wavelets and no brush that works for me, but I was intrigued by the idea of just subtracting the blue channel to “erase” just a tad of the bad details in the skin. But after doing that I lacked contrast, so the idea arose to bring that back using another channel, the red. which would not bring back the bad details.
Thanks for posting this. It’s in GIMP yes? I intend to start using GIMP as I used to like layers in P’shop. At this stage I can only guess what overlay mode and grain extract are, but I have some portrait photos waiting to be processed (RT then GIMP I expect) and your technique looks useful. It’s brilliant what appears in these forums.
I’m curious what you’ll come up with. Furthermore I just discovered your B&W conversion tuts and your script fu for all the color channels…that obviously increases the possibilities a lot…
