基于高阶张量的多属性压缩融合体绘制方法研究

发布时间：2018-06-10 02:45

本文选题：多属性体数据 + 四阶张量近似　；参考：《电子科技大学》2017年硕士论文

【摘要】：随着数据规模的越来越大,压缩体绘制是解决大规模数据体绘制的有效方法。在众多压缩体绘制方法中,基于张量近似的压缩体绘制方法在压缩率和绘制效果方面具有良好的性能。多属性融合可视化是降低目标多解性、凸显目标特征的有效手段。本文针对多属性压缩体绘制问题开展研究,并取得了一定的成果。该研究具有较大的理论价值和实际应用价值。针对多属性融合压缩体绘制问题,在现有三阶张量近似的压缩体绘制和多属性融合体绘制的基础上,探索四阶张量近似的压缩体绘制方法。同时,针对张量近似中的秩截断对压缩率和压缩效果有较大影响问题开展研究。主要的工作和创新如下:1.提出了基于四阶张量近似的多属性压缩融合体绘制方法。针对多属性压缩融合体绘制问题,目前还没有有效的方法和手段,但可以利用三阶张量近似分别对多个属性体数据进行压缩,并在体绘制阶段进行融合绘制。这种方法压缩效率较低,无法充分利用多属性数据之间的相关性。基于此,本文提出了基于四阶张量近似的多属性融合体绘制方法。其基本思想是将多属性体看作一个四维数据并进行基于四阶张量近似的压缩、融合。仿真结果表明,本方法在损失极少的数据信息的情况下,数据的压缩率有明显地提升;2.提出了基于各向异性秩截断的高阶张量近似方法。在高阶张量近似中,通常是各个方向上的采用相同的秩。而实际数据往往在不同方向上的相关性具有各向异性的特点。因此本文提出了基于各向异性秩截断的高阶张量近似方法。该方法通过对各个方向上奇异值分布的分析,计算各个方向的秩截断大小,得到不同方向上的截断秩,以提高压缩效率和压缩效果。仿真结果表明,本文提出的基于各向异性秩截断的高阶张量近似方法在满足压缩率的情况下,压缩效果有所提升。综上所述,本文在仿真平台上实现上述方法。仿真结果表明,本文提出的方法在多属性体数据压缩融合体绘制中有效提高了数据的压缩率。在针对各向异性数据张量近似截断秩选取的问题中,本文提出的方法在不同压缩率限定条件下有效提高数据的压缩效果,在多属性数据的可视化中有着很好的应用前景。
[Abstract]:With the increasing scale of data, compressed volume rendering is an effective method to solve large scale data volume rendering. Among many compressed volume rendering methods, the compressed volume rendering method based on Zhang Liang approximation has good performance in compression ratio and rendering effect. Multi-attribute fusion visualization is an effective method to reduce the multi-solvability of targets and highlight the features of targets. In this paper, the problem of multi-attribute compression volume rendering is studied, and some achievements have been made. This research has great theoretical value and practical application value. Aiming at the problem of multi-attribute fusion compression volume rendering, based on the existing third order Zhang Liang approximation compression volume rendering and multi-attribute fusion volume rendering, the fourth order Zhang Liang approximation compression volume rendering method is explored. At the same time, the influence of rank truncation on compression ratio and compression effect in Zhang Liang approximation is studied. The main tasks and innovations are as follows: 1. A method of multi-attribute compression fusion volume rendering based on fourth order Zhang Liang approximation is proposed. There are no effective methods and methods for multi-attribute compression fusion volume rendering, but it is possible to compress multiple attribute volume data using third-order Zhang Liang approximation and perform fusion rendering in volume rendering phase. The compression efficiency of this method is low and the correlation between multi-attribute data can not be fully utilized. Based on this, a multi-attribute fusion volume rendering method based on four-order Zhang Liang approximation is proposed. The basic idea is to treat the multi-attribute body as a four-dimensional data and to compress and fuse it based on the fourth-order Zhang Liang approximation. The simulation results show that the compression ratio of the data can be significantly improved when the data information is lost very little. A high order Zhang Liang approximation method based on anisotropic rank truncation is proposed. In higher order Zhang Liang approximation, the same rank is usually used in each direction. The correlation of actual data in different directions is anisotropic. Therefore, a high order Zhang Liang approximation method based on anisotropic rank truncation is proposed. By analyzing the distribution of singular values in different directions, the rank truncation of each direction is calculated, and the truncation rank in different directions is obtained to improve the compression efficiency and compression effect. The simulation results show that the proposed high order Zhang Liang approximation method based on anisotropic rank truncation can improve the compression performance when the compression ratio is satisfied. To sum up, this paper implements the above method on the simulation platform. Simulation results show that the proposed method can effectively improve the compression rate of multi-attribute volume data compression fusion volume rendering. In order to solve the problem of selecting approximate truncation rank of Zhang Liang for anisotropic data, the method proposed in this paper can effectively improve the compression effect of data under different compression ratios, and has a good application prospect in the visualization of multi-attribute data.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;O183.2

【相似文献】