Abstract: In a process for rendering a 3-dimensional graphics scene made up of a plurality of static and/or dynamic objects composed of geometrical elements, a method for obviating redundant processing and communications involved with hidden dynamic objects, so as to accelerate the rendering of successive images of the scene and to reduce communications requirements (FIG. 6C). The method includes in respect of each occluded dynamic object: (a) predicting a time period during which the object is expected to remain occluded; (b) generating a volume that contains the object during the time period; (c) inserting the volume into a spatial data structure used by an occlusion culling technique; (d) applying the occlusion culling technique so as to output-sensitively render the scene, yet not rendering or updating the occluded dynamic object during the time period, provided the occlusion culling technique does not find the volume to be visible.