Jan Fischer's Homepage - VMV 2005 Paper Details

 Fast Rendering of Large Encoded Isosurfaces from Uniform Grid Datasets


Standard algorithms for the extraction of isosurfaces from volume data (e. g., the Marching Cubes algorithm) are notorious for producing a large amount of small triangles. Unfortunately, simplification is not a possible avenue in many application fields, thus a large number of triangles are transferred to the GPU. The resulting massive data load of transferring polygonal data from main memory down to the GPU is a major bottleneck in traditional polygonal isosurface rendering. In this paper we propose a novel solution for rendering large isosurface models. Our approach is based on the combination of an encoding scheme for the isosurface and a hardware-accelerated decoding strategy that takes place entirely on the GPU. The encoding scheme takes into account the underlying regular grid structure inherent to this type of surfaces, hence it results in a more compact representation. By using this (lossless) compressed version of the polygonal model, we reduce both the communication between CPU and GPU and the storage requirements without any appreciable loss of quality. Furthermore, no further (slow) bus transfers are necessary, since decoding is performed on the GPU and the decoded geometry is written directly into on-board graphics memory. Therefore a significant speeding up of the overall rendering is achieved. An application example is presented in order to illustrate the benefits introduced by our approach.


   AUTHOR = {A. del R{\'i}o and J. Fischer and D. Bartz and W. Stra{\ss}er},
   TITLE = {{Fast Rendering of Large Encoded Isosurfaces from Uniform Grid Datasets}},
   BOOKTITLE = {{Vision, Modeling, and Visualization (VMV)}},
   MONTH = {November},
   YEAR = {2005},
   PAGES = {71--78},
   LOCATION = {Erlangen}


Document  Download paper [PDF]