I’ve just found back some links explaining a grayscale conversion method named after Greg Gorman and Mac Holbert, and also called the Carr method.
It is a conversion method based on the CIELab L channel, and which is supposed to give very good results on portraits.
The steps are quite obscure, but the results are sometimes really nice!
Here are couple of links explaining how to realise the method in photoshop and ImageMagick, but I suppose the same can be done with GIMP with minor modifications…
Wow, that was a blast from the past. I think I had looked at this years ago as a B&W conversion method. I got it to work, but don’t remember the steps I did back then. I suppose we can have another look at it.
[update]
Ok, I’ve just walked back through the steps in GIMP…
Colors → Components → Decompose
Color model: LAB
Uncheck Decompose to layers.
This gives you three new images, L, a, and b. Delete a and b. Work on the L image.
On the Channels tab, right-click the Gray channel → Channel to Selection.
Invert selection (Selection → Invert). (So you’re selecting the shadows).
Go back to an RGB image (Image → Mode: RGB).
Duplicate the Base layer.
On the duplicate layer, fill it with your chosen fill color (Black for a straight B&W conversion). This will fill in the selection (shadows).
Change layer blend mode to Multiply.
Make a new layer from visible.
Change it’s blend mode to Overlay, adjust as desired (20% in your linked example).
