The goal of this project is to create a software application that permits the visualizationof 3D sampled scalar data, also known as Volume Data, using a technique calledDirect Volume Rendering (DVR). This technique allows the creation of an image thatrepresents the entire data set, including the visualization of the interior of the volume,i.e., not just the surface. The drawback of this technique is the tremendous storageand processing requirement, since it needs the entire volume information in order togenerate the image. In order to generate this image one of the DVR algorithms waschosen: the Ray-Casting algorithm.
The software allows the user to interact with the volume, changing the colour andthe opacity, the illumination and rotation, among other things. Volume rotations canbe done in real-time, the other changes are incorporated in a pre-processing step.To be able to achieve real-time volume rotations, the Ray-Casting algorithm wasoptimized and parallelized. To speed up even more the frame rate, a different versionof the parallel algorithm was developed, called the interactive mode, which sacrificessome image quality.
he Message Passing Interface (MPI) was used as a tool for the parallelization.The software was developed and tested on the Beowulf cluster “Calhau”. This clusterallows a real-time rotation of large volumes, up to 256^3 voxels.
2002. , 50 p.