Umfassende visuelle Informationssuche in multidimensionalen Datensätzen

该项目扩展了穷举视觉搜索方法的结果，以解决多维数据集中的高阶关系，这些数据集将通过将图像分析技术应用于大量自动创建的可视化来进行搜索，建模和分析。在这里，“高阶关系”是指考虑由用户绘制的草图描述的两个维度之间的非平凡关系、两个以上维度之间的关系以及仅存在于连续（非离散）数据集中的关系。对于所有三种情况，都应探讨基于详尽视觉搜索的解决方案。从方法上讲，这些方法基于为不同的可视化找到新的质量度量，分析3D可视化，以及在连续可视化中搜索特征。在专注于开发通用（独立于应用程序）原则的同时，我们的方法将在我们的气候研究和医学成像合作伙伴提供的高维和大型数据集上进行测试。

The project extends the results of the exhaustive visual search approach to address higher-order relations in multidimensional data sets which are to be searched, modeled, and analyzed by applying image analysis techniques to a large number of automatically created visualizations. Here, “higher-order relations“ refers to considering non-trivial relations between two dimensions described by user-drawn sketches, relations between more than two dimensions, and relations only present in continuous (nondiscrete) data sets. For all three cases, solutions based on the exhaustive visual search shall be explored. Methodically, the approaches are based on finding new quality measures for different visualizations, analyzing 3D visualizations, and searching features in continuous visualizations. While focusing on developing general (application-independent) principles, our methods will be tested on high-dimensional and large data sets provided by our cooperation partners from climate research and medical imaging.