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.