Implicit Ray Casting of the Parallel Vectors Operator

Feature extraction is an essential aspect of scientific data analysis, as it allows for a data reduction onto relevant structures. The extraction of such features from scalar and vector fields, however, can be computationally expensive and numerically challenging. In this paper, we concentrate on 3D line features in vector fields that are defined by the parallel vectors operator. Common examples are vortex corelines and hyperbolic trajectories, i.e., lines around which particles are rotating, or from which particles are repelled and attracted locally the strongest. In our work, we use a GPU volume rendering framework to calculate the lines on-the-fly via a parallel vectors implementation in the volume rendering kernels. We achieve real-time performance for the feature curve extraction, which enables interactive filtering and parameter adjustment.