Acevedo Daniel, Laidlaw David H
Department of Computer Science, Brown University, USA.
IEEE Trans Vis Comput Graph. 2006 Sep-Oct;12(5):1133-40. doi: 10.1109/TVCG.2006.180.
We present an evaluation of a parameterized set of 2D icon-based visualization methods where we quantified how perceptual interactions among visual elements affect effective data exploration. During the experiment, subjects quantified three different design factors for each method: the spatial resolution it could represent, the number of data values it could display at each point, and the degree to which it is visually linear. The class of visualization methods includes Poisson-disk distributed icons where icon size, icon spacing, and icon brightness can be set to a constant or coupled to data values from a 2D scalar field. By only coupling one of those visual components to data, we measured filtering interference for all three design factors. Filtering interference characterizes how different levels of the constant visual elements affect the evaluation of the data-coupled element. Our novel experimental methodology allowed us to generalize this perceptual information, gathered using ad-hoc artificial datasets, onto quantitative rules for visualizing real scientific datasets. This work also provides a framework for evaluating visualizations of multi-valued data that incorporate additional visual cues, such as icon orientation or color.
我们展示了对一组基于二维图标的参数化可视化方法的评估，在此过程中，我们量化了视觉元素之间的感知交互如何影响有效的数据探索。在实验中，受试者针对每种方法量化了三个不同的设计因素：它能够表示的空间分辨率、它在每个点能够显示的数据值数量，以及它在视觉上的线性程度。该可视化方法类别包括泊松盘分布图标，其中图标大小、图标间距和图标亮度可以设置为常数，或者与二维标量场中的数据值相关联。通过仅将这些视觉组件之一与数据相关联，我们针对所有三个设计因素测量了过滤干扰。过滤干扰表征了恒定视觉元素的不同级别如何影响与数据相关联元素的评估。我们新颖的实验方法使我们能够将使用临时人工数据集收集的这种感知信息推广到用于可视化真实科学数据集的定量规则上。这项工作还提供了一个框架，用于评估包含其他视觉线索（如图标方向或颜色）的多值数据的可视化。
