Real-time interactive time correction on the GPU

The study of physical phenomena and their dynamic evolution is supported by the analysis and visualization of time-enabled data. In many applications, available data are sparsely distributed in the space-time domain, which leads to incomprehensible visualizations. We present an interactive approach for the dynamic tracking and visualization of measured data particles through advection in a simulated flow. We introduce a fully GPU-based technique for efficient spatio-temporal interpolation, using a kd-tree forest for acceleration. As the user interacts with the system using a time slider, particle positions are reconstructed for the time selected by the user. Our results show that the proposed technique achieves highly accurate parallel tracking for thousands of particles. The rendering performance is mainly affected by the size of the query set.