If you think of a head with a headband, the headband has 4 points. The head moves through a range of motion, I want to generate the pivot location/orientation in every frame from these moving points.
If you've got the points in 3D space, it should just be an application of trigonometry to compute the position and orientation of the parent body. Or are you getting the the points in 2D space (like objects in a video) and you are trying to find positions in 3D?