Advanced Quality Metrics in Information Visualization and Scientific Visualization

对于高维数据的可视化自动分析，质量度量是一个很有前途的概念。为了完全可视化高维数据集，需要进行大量的可视化。其中只有一部分显示了数据的相关结构，因此用户只需要看到其中的一部分。质量度量的思想是自动检测良好可视化的这个子集。为此，它们模仿视觉感知系统。已知的几个质量度量主要针对高维数据的双变量离散可视化分析。该项目在三个方面扩展了传统的质量度量方法。质量度量的概念将被扩展到多元投影中的非线性嵌入，它将被扩展到非离散可视化(又名连续可视化)，并且它将被扩展以度量质量度量的可靠性。由于概念扩展是相互关联的，我们建议在一个项目中处理它们。

Quality metrics are a promising concept for the automatic analysis of visualizations from high-dimensional data. In order to completely visualize a high-dimensional data set, a large number of visualizations are required. Just a subset of them shows relevant structures of the data and thus just a subset of them is required to be seen by the user. The idea of quality metrics is to automatically detect this subset of good visualizations. For this, they mimic the visual perception system. A couple of quality metrics are known aiming mostly at the analysis of bi-variate discrete visualizations of high-dimensional data. This project extends the traditional approach of quality metrics in three ways. The concept of quality metrics will be extended to nonlinear embeddings in multivariate projections, it will be extended to non-discrete visualizations (aka continuous visualizations), and it will be extended in order to measure the reliability of quality metrics. As the conceptional extensions are mutually related, we propose their treatment within one project.