My process: take the original image, duplicate and set it to screen and then do over a dozen adjustment layers to get the coloring right. Do noise reduction before cropping/resizing and then unsharp mask after. Then use those adjustment layers for the other screencaps *
I've also been using linear burn/multiply and color dodge layers as demonstrated by
sinspire over here. They work especially well on uniforms/clothing - like for Roy in this icon and for Ed over hereAnd I couldn't stop smiling like an idiot at Ed in this pic bellow - was very happy with how much color I was able to get using just adjustment layers :3
to 
(Click to zoom)
EDWARD IS SUCH A CUTIE, OMG. ( Now the name no longer makes me think of Twilight, thank god. Alchemists all the way baby, \o/ )
Still, I find icon making takes forever and ultimately time seems to go by faster when I'm doing CSS >_> I can't churn out hundreds of icons every month, especially when its so much easier to stalk other icon makers, lol.
* Though if anyone is interested in the technical mumbojumbo, what I did was while doing my first icon I created an action which recorded all my steps. Then I used the automated batch processing to go through the 50+ caps and run the action on them. Then I just tweaked the layers accordingly for each of the new screencap PSD files.That way I could get filters such as smart sharpen applied automatically too :3
