Incremental Slicing Revisited : Accelerated Volume Rendering Of Unstructured Meshes

We accelerate the incremental slicing method for near interactive volume rendering of unstructured grids with conventional polygon rendering hardware support. Our method uses less memory compared to the method of Yagel et al. [25], with significant improvements in rendering time. Using our method, a rendering performance gain of at least 400% was achieved for some datasets, e.g., the Blunt fin dataset. We completely eliminate sorting of edges as a preprocessing step. Further, we introduce a hierarchical z-region paradigm, which can be incorporated with the dataset-input module to get the depth information of the cells with no extra computational overhead. We argue that keeping an active edge list and an active cell list as data structures is redundant, because this information is inherent in the z-region paradigm in our method. In pursuit of accuracy, we explore the application of Nyquist’ s classical sampling theorem when slicing the datasets, and conclude that images of comparable visual quality can be obtained by relaxing this requirement, while maintaining the rendering performance. Finally, we conclude with the results of our method, and a discussion on how progressive refinement of 2-manifold triangular meshes (for 2D slices of cutting planes parallel to the view plane) can help to achieve additional interactivity in a volume visualization system. CR Categories: I.3.5 [Computer Graphics]: Computational Geometry and Object Modeling – surfaces and object representations.