Institute for Data Analysis and Visualization

We present three enchancements to accelerate the extraction of separatrices of three-dimensional vector fields, using intelligently selected ''sample'' streamlines. These enhancements reduce the number of needed sample streamlines and their propagation length. Inflow/outflow matching supports the simultaneous extraction of topologically significant inflow and outflow separatrices in a single pass. An adaptive sampling approach is introduced and used to seed streamlines in a more meaningful and efficient manner. Cell-locking is a new concept that isolates regions of a data set that do not contain separatrices. This concept makes steamline propagation more efficient as streamlines are not propagated through cells that do not influence or contain separatrices. These enhancements enable us to perform separatrix construction for three-dimensional vector field data requiring less overall computation.