Interactive volume rendering based on ray-casting for multi-core architectures

The Volume Ray-Casting rendering algorithm, often used to produce medical imaging, is a well-known algorithm and the underlying computation can be easily executed in parallel. This is due to the fact that the huge number of rays, used to sample the volumetric data, can be processed independently. However, the algorithm’s performance may drop substantially when the complexity/size of the volumetric dataset increases. In this paper, we present three implementations of our parallel volume ray-casting algorithm in different multi-core architectures, such as CMPs, GPUs and MPSoCs. Furthermore, we show that using multi-GPUs, that perform in parallel, we can almost halve the rendering time. The performance and aspects of the three implementations are discussed