I have read on the web of a technique that recovers (at least to some degree) edge detail that may be lost in green screen keying when the matte is being scaled and processed. It supposedly works by sandwiching a copy of the original FG plate between the keyed version above and the BG below. That's as far as the description goes. If indeed this is so, I'm guessing some combination of color modification, blending modes, and opacity come into play. But so far, my experimentation hasn't produced the desired results.
I regularly use something kind of like this in sky replacement where fine detail like telephone wires is lost but can be recovered by placing the original sky footage over the sky replaced footage and reducing it's opacity enough to see both the new sky and the fine detail.
Anyway, does anyone have any insights to what I'm trying to figure out?
From your description, it sounds like Light Wrap. It's a process that more convincingly incorporates pixels from the keyed object into the background.
It's a very subtle thing: you can't overcome obvious mistakes like the wrong angle for the key light, color balance, color grading, etc. These are problems created at image acquisition, which Light Wrap is unable to know about.
There are third-party plugins that do Light Wrap (the Sapphire suite, for one) and it can be done for free by jumping through a lot of hoops in AE.
Former Sr. Promotion Producer
KCRG-TV (ABC) Cedar Rapids, IA
Thanks to both. I'm familiar with light wrap and have rolled my own as well as used a script from AE enhancers. I need to think this through though honestly, I'm not sure. I dug through my stack of notes and found the reference in a description of a training program from cmivfx.com under Nuke advanced keying. No details are given at all and they include light wrap as a separate topic.