Real-time ray casting architectures for volume graphics and volume rendering

Real-time visualization of large volume datasets demands high performance computation, pushing the storage, processing, and data communication requirements to the limits of current technology. Volume visualization using the ray casting algorithm involves the projection of a large 3D array of data (voxels) onto a 2D array of pixels multiple times per second. General purpose parallel processors have been used to render moderate size datasets at interactive frame rates; however, the cost and size of these approaches inhibit the widespread use for real-time visualization. Custom approaches have promised low-cost high-performance for the personal computer. Most recently, a custom ASIC, VolumePro, has been developed by Mitsubishi Electronic Research Laboratory. Specialized solutions tend to have cost, size, and performance advantages over their predecessor. Therefore, we believe that they will emerge as the dominant means of real-time visualization. Furthermore, it is likely that personal computers will be the platform of choice because of their increased computational power and availability. This thesis improves upon first generation specialized approaches by increasing the rendering efficiency, enabling real-time perspective, and stereoscopic rendering for virtual reality displays. Interactive volume visualization is dominated by O( n3) computational and memory throughput. We improve performance over current solutions by using a hybrid approach. As a result, this ray casting architecture will allow us to interactively visualize next generation ≥5123 datasets with similar resources that first generation volume rendering approaches use to render ≤2563 datasets. Furthermore, by efficiently accessing and rendering the dataset in small-blocks, we can potentially off-load the dataset into the host processor for low-cost implementations such as an accelerated graphics port (AGP) on desktop computers.