This paper proposes some methods for accelerating ray tracing which is a widely used technique to obtain realistic images but very time-consuming process mainly due to the large amount of intersection calculations. Parallelization of the algorithm seems to be the unique or, at least, most promising solution for significant seed-up.
In chapter 3, it is proposed that a new parallel ray tracing hardware architecture where processors are connected as a ring. Most existing parallel ray tracing algorithms subdivide whole object space into subregions and assign each subregion to each processor which then handles only rays entering the subregion associated with it. Here each processor is generally assigned objects not confined to a local subregion but rather spread over the whole object space. Each processor individually executes ray tracing on its objects to obtain the respective partial result, which is then composed with each other to form the final image. The proposed scheme is especially suitable for synthesizing animated sequence of images because time-consuming object reallocation process needs not to be performed at the start of every frame which is not the case in other algorithms. Preliminary results show that the proposed architecture exhibits a speed-up factor almost linearly proportional to the number of processors. This paper also show that the proposed system can accelerate scanline rendering process so that the propose system can be used in real-time applications.
In chapter 4, this paper proposes a new method for computer animation which eliminates the viewing transformation for each frame as required in the conventional method. For keyframe animation systems, the proposed method has been analytically and experimentally shown to be effective if the number of in-between frames is greater than 7 in the worst case. It can also be applied to other animation systems such as dynamic simulation system. And then it also proposes some methods of using frame coherence for ray tracing, thereby reducing the computing time.
In chapter 5, this paper proposes a new accelerating method for ray tracing in a network environment which consists of various computers. Firstly, it describes some key-points for network rendering and proposes a new method satisfying these key-points. The propose method uses without dataflow model and token for dynamic load balancing and prevention of network contention which is a frequent situation in a server-client model.
