Large displays and hand gestures can be combined to augment the traditional desktop paradigm, allowing for multiple co-located users to interact simultaneously with a single computer, and therefore to engage in computer mediated collaboration activities. Here we introduce a novel technique for recovering the rotation of the hands, adapting a previously developed vision-based interface which we use to extract the users' bare hands from a video stream, taken over a tabletop display. Unlike previous tabletop input devices, ours is focused on capturing holistic hand cues. We argue that these cues, namely the posture and the rotation of the hand, can be an important complement to the fingertips' positions used in most systems of this kind, because we consider they can make the experience more natural and simple for the users. We also present a new interaction technique using these capabilities.