Three-dimensional position information is used to segment objects in a
scene viewed by a three dimensional camera. At one or more instances of
an interval, the head location of the user is determined. Object-based
compression schemes are applied on the segmented objects and the detected
head.