Compression of multiphase Volume-of-Fluid datasets

Abstract

To visualize Volume of Fluid (VOF) datasets, it is necessary to handle the large VOF dataset files, which is slow. This work presents a method for compressing multiphase VOF datasets using a hierarchical tree data structure and a Level of Detail (LOD) approach for visualizing the VOF data. The data structure forms a sparse representation of the VOF field. This way, it makes use of the spacial coherency of VOF datasets to achieve high compression rates. A custom file format for storing the hierarchical data structure is presented. The file format is designed in a way that allows reading the file only partially to obtain a lower resolution version of the data. The LOD approach is used to quickly render an image of the data stored in this format for visualization purposes. Piecewise Linear Interface Calculation (PLIC) is used to extract the fluid surfaces from the VOF data. The system uses parallel processing to reduce computation times. The hierarchical data structure of the compression method is used to generate the various levels of detail for the LOD approach. In addition to the implementation of the proposed methods, a large set of benchmarks are used to test both the compression algorithm and the visualization system. The compression algorithm is compared against the performance of the more common generic compression methods LZ4, zlib and the zip file format. Both the compression time and compression rate were assessed. Especially in cases where the dataset exhibits strong spatial coherency the proposed method can achieve better compression rates than the other compression methods. For the visualization system, the rendering time was measured. The calculation of the PLIC surfaces in the visualization system is tested on both the CPU and the GPU. Both implementations can achieve fast rendering times.