A Taxonomy for Feature Tracking in Time-Varying Piecewise-Linear Scalar Fields

This paper introduces a taxonomy that unifies a broad class of feature tracking approaches for time-varying, piecewise-linear scalar fields in scientific visualization, and describes its realization as a pipeline-based processing framework. The lack of a universal feature definition or tracking approach suitable for every application domain has led to the development of numerous specialized, monolithic methods that target specific feature types, rely on distinct similarity metrics, employ customized correspondence-assignment optimization schemes, or utilize different visualization techniques. Based on a survey of several state-of-the-art tracking methods, we first derived a taxonomy capable of categorizing these techniques and then decomposed the methods into subprocedures, which were reimplemented as reusable, interchangeable modules within the framework. We implemented this framework in the widely used Topology ToolKit (TTK) and Visualization ToolKit (VTK) libraries, which can be accessed via C++, Python, or the ParaView graphical user interface, thereby reaching a broad community of researchers and practitioners. We demonstrate the value of the taxonomy and its accompanying framework by replicating results from the surveyed publications, and by preliminarily exploring the composition, comparison, and validation of different tracking approaches, which the framework enables.