There are also parallel rendering algorithms based on raytracing and raycasting, which follow individual light rays to or from the viewer. These are inherently parallel (each ray is independent), however they require a copy of the entire dataset on each node, or else lot of costly data movement.
NASA JPL has developed a parallel ``ray-identification'' program using a method of this kind. This independently transforms pixels into the view plane, and uses Z-buffering for hidden surface removal. Scattered data decomposition is used for load balancing.
These methods produce high-quality results but are very computationally intensive. The JPL program has been used to make movies of simulated flights over Venus, Mars, and the Earth. It can be run interactively (1.5 fps) on a 256-node Intel Paragon.