Remote user-driven exploration of large scale volume data

No rendering pipeline exists to explore very large volume data, for example on the order of terabytes or more. In the extreme case, the data is essentially “stuck” at the site of creation. The consequence is that the utility of the dataset is greatly diminished since only a very small number of users are able to explore the rich information contained within. We address the challenge of how to provide access to such datasets to remote user's equipped with low-cost computational and display technology. Our work proposes a novel end-to-end rendering pipeline that allows for effective data exploration. Our paradigm couples view-dependent and image-based data structures along with novel rendering algorithms that allow for fast spatial and transfer function browsing on the client side. The view-dependent data structure, called a Pixel Ray Image (PRI), holds scalar information on projection rays through the volume. The representations for the scalar data are determined from the requirements of the particular projection equation. We present compact representations that may be stored in texture form for interactive rendering on today's PC graphics hardware. During spatial browsing, when the transfer functions do not change, the Layered Slab Image (LSI) data structure holds pre-computed projections of the data that are then used to quickly compute approximate renderings at nearby view-points. The PRI represents a compression of the volume along a single dimension. Sampling planes placed orthogonal to the sampling direction of the PRI can be computed easily in graphics hardware. Using an approach similar to the shear-warp algorithm, the volume may be rendered directly from this compressed format at views within a 45 degree neighborhood. The volume may be rendered from any view by placing three PRI at orthogonal sampling directions to each other. We present a novel rendering algorithm to render the volume directly from the compressed format on PC graphics hardware.