I've been working on manipulating stereo images for about a year, and recently shot several images in Rocky Mountain National Park that Served My Needs. There's an album with 10 images in it most from there, that I would very much appreciate comments on. (Though since I'm away from the computer a lot, I may not respond speedily. My apologies in advance.)
Here's an example (not from RMNP):
Evening falls on the bay.
These turn out to be not-very-complicated to do. All of them are taken using the sway method with a non-stereo camera (take a shot, lean to the right, take another shot). I've also heard it called the "Cha-cha" technique, but when I cha-cha there is a lot of hip movement (an alarming amont, I've been told), but taking these, my feet are firmly planted, and I sway from the waist.
To make the images I take corresponding segments of the right and left singles of a stereo pair. In Photoshop I use Image/Image rotation/Flip Canvas Horizontal on (usually) the right hand image. Then I line it up with the left half. This is the fiddly bit, because things at different distances from the camera will be further right or left. So I line up the most distant areas so they are as mirror-imaged as possible then make my way down erasing parts of the right hand image: more erasure for the close-up portions, less for the further-away portions.
I look at them at nearly max magnification in PS to join and erase. Tedious. If all has gone well (and you will notice it didn't in the image above) everything will be a fairly close mirror image along the join of the two parts of the original pair. The result will have made the left half of the new pair.
My one real insight into this was to realize that once you have half of the pair, the other half of the pair is very easily made; you just flatten your layers and copy all. Open a new image (File/New in PS), paste the copy in and flip it (Image/Image rotation/Flip Canvas Horizontal). That completes the right hand half of the new pair. Copy that and paste it alongside the left-hand half and there you go!
One of the things I like about these manipulated pairs is that it is super-easy to line up the height of the pair to the exact pixel. Once you have both sides, their edges will join perfectly, since one is a flipped version of the other. So if you look at the join between the two halves of the pair above, it is much more precisely in-line than the join within each image.
Give it a try, as the pusher said, it's fun!