Neophytos Neophytou - Irregular Splatting
GPU-Accelerated Volume Splatting With Elliptical RBFs
Neophytos Neophytou and Klaus Mueller
In collaboration with Kevin T. McDonnell, Wei Hong, Xin Guan, Hong Qin, Arie E. Kaufman
 |
 |
|---|---|
 |
 |
Abstract:
Irregular grids still represent the most challenging discrete dataset topology. This paper focuses on the rendering aspects of these. Traditionally, the renderer of choice has been the cell-based approach devised by Shirley and Tuchman, and a variety of GPUaccelerated implementations have followed this pioneering work. In this paper, we take an alternative view, considering the irregular data as scattered data points, and a natural rendering paradigm for these is Splatting. However, the extension of splatting to irregular grids poses several new challenges, some of which are shared with cell-based approaches. These are the need for proper depth ordering and gradient estimation as well as the prevention of blurring artifacts. As a solution we employ a post-shaded imagealigned slicing approach and accelerate it on the GPU for speed. In order to reach a maximum of efficiency, we exploit various GPU features, such as early z-culling, to restrict fragment operations to only those data points and screen areas that are relevant at the current viewpoint and transfer function setting. We demonstrate our renderer using datasets from subdivision volumes and computational science, elaborately transformed to work optimally with Splatting.
Publications:
- N. Neophytou, K. Mueller, K. T. McDonnell, W. Hong, X. Guan, H. Qin and A. Kaufman,
GPU-Accelerated Volume Splatting With Elliptical RBFs,
Joint Eurographics - IEEE TCVG Symposium on Visualization 2006,
Lisbon, Portugal, May, 2006.
- W. Hong, N. Neophytou, K. Mueller, and A. Kaufman
Constructing 3D Elliptical Gaussians for Irregular Data,
to appear in: Moeller, T., Hamann, B. and Russell, R.D., eds.,
Mathematical Foundations of Scientific Visualization, Computer Graphics,
and Massive Data Exploration, Springer-Verlag, Heidelberg, Germany 2006.
Implementation Notes:
We extend the GPU framework described in previous work to enable the visualization of irregular datasets. The basic primitive of a spherical Gaussian kernel is now extended to a general ellipsoid. Our new framework is able to interactively slice ellipsoidal kernels of varying size and orientations.
More details on the geometry of slicing ellipsoids as well as a description of our GPU implementation will follow in the next few weeks.
Go back to Research