Rotoscoping is simple once you understand the process.  You take one or more video views of a particular motion and place them into your 3D scene (as background animation or textured onto cubes/planes).  You then must line up the target figure to overlap the figure in the video(s) and you start posing to match the video - one keyframe at a time.

The problem with rotoscoping - especially from one video view - is that you don't have all of the information - unless you recorded it yourself and planned well ahead (markers like for motion capture systems).  It can be difficult to determine what rotations/translations represent the actual orientations/positions of the video actions.  You'll need to understand anatomical dynamics to avoid awkward poses and you'll need to check orthogonal views to avoid them as well.

It is painstaking work, but the results are far more realistic than simple animation but not as good as quality motion capture.  One advantage over motion capture is that you can do this with interpolated keyframes instead of time-sampled keys (which take up oodles of memory).

Robert

C makes it easy to shoot yourself in the foot. C++ makes it harder, but when you do, you blow your whole leg off.

 -- Bjarne Stroustrup

Contact Me | Kuroyume's DevelopmentZone