Compositional Analysis and Synthesis of Scientific Data Visualization Techniques

Abstract

The scientific data visualization process involves a sequence of transformations that convert a data set into a displayable image. One of the most important transformations in this process is the visualization mapping which defines a set of bindings between data and graphical primitives. Since these bindings describe how data is going to be visualized, the effectiveness of visualization critically depends on the mapping defined at this stage. Establishing a proper mapping which leads to an effective data visualization requires significant knowledge in several fields, such as, data management, computer graphics, and visual perception. However, scientists who could benefit most from data visualization usually lack this knowledge. In order to identify, acquire, formalize, and provide this knowledge, the existing visualization techniques, that are known to be useful, have been thoroughly analyzed. The analysis shows that most of the existing data visualization techniques can be described in terms of attributes of data, a set of primitive visualization techniques, marks (graphical symbols) that modify primitive visualization techniques, and a set of rules used in their design. The analysis further suggests a design process leading to the automatic synthesis of scientific data visualization techniques.