The Flipping Cube: A Device for Rotating 3D Rasters

Driven by the prospect of three-dimensional rasters as a primary vehicle for future 3D graphics and volumetric imaging, this paper introduces an architecture for real-time rendering of high-resolution volumetric images. The Flipping Cube Architecture utilizes parallel memory organization and a unique data orientation scheme in order to support contention free access to viewing rays. The swift advances in performance, availability and price of computing power, memory, and disk-space are transforming long thought techniques into reality. One typical example to this trend is the revolution taking place in the eld of volume graphics. Grasping the feeling of this revolution requires no more than three decades of perspective. The display of computer graphics in the sixties was based on vector drawing devices and an 'object based' approach to scene representation and rendering. A symbolic representation of the scene objects was stored in a display-list and managed by the computer. Refreshing the screen was accomplished by rendering the vectors comprising the objects in the display-list. An alternative approach, termed raster graphics utilizes a 2D frame-buuer of pixels for scene representation and a 'raster based' renderer. This renderer colors the pixels in the frame-buuer that correspond to the discrete representation of the symbolic objects. Screen-refresh is performed by re-displaying the frame-buuer on the screen, thus there is no need to save the symbolic object representation. The main disadvantage of this approach is the memory and processing power it requires, which delayed its appearance until the early seventies when the technology was able to provide cheap and fast hardware to support it. The major advantages of vector-graphics are its ability to perform object related operations on the display-list and the fact that the lines it draws are continuous (i.e., without aliasing). On the other hand, the main appeal of raster-graphics is that it decouples rendering from screen-refresh which makes it insensitive to the scene complexity. It is also able to perform block operations, and it is suitable for the display of sampled images, that is, graphics and digitized images can be intermixed. Moreover, the 'object based' approach can still be imitated by maintaining a display-list which is redrawn to the frame-buuer as a result of each change in the scene. The above discussion describes the state-of-the-art in 2D graphics. In 3D graphics, however, the ruling majority of rendering methods still employ an 'object based' approach called surface graphics rather than employing a three dimensional 'raster …