Advances in massive model visualization in the CYBERSAR project

We provide a survey of the major results obtained within the CYBERSAR project in the area of massive data visualization. Despite the impressive improvements in graphics and computational hardware performance, interactive visualization of massive models still remains a challenging problem. To address this problem, we developed methods that exploit the programmability of latest generation graphics hardware, and combine coarse-grained multiresolution models, chunk-based data management with compression, incremental view-dependent level-of-detail selection, and visibility culling. The models that can be interactively rendered with our methods range from multi-gigabyte-sized datasets for general 3D meshes or scalar volumes, to terabyte-sized datasets in the restricted 2.5D case of digital terrain models. Such a performance enables novel ways of exploring massive datasets. In particular, we have demonstrated the capability of driving innovative light field displays able of giving multiple freely moving naked-eye viewers the illusion of seeing and manipulating massive 3D objects with continuous viewer-independent parallax.