Jan Dvořák

We propose improved center affinity based on maximum value encountered up to the currently processed frame. We also detect and filter irregularities and perform a global optimisation in post processing to further improve the tracking quality.

In this work, we propose a priority-driven approach for encoding mesh connectivity, if the geometry is already known (e.g., encoded separately using inter-frame prediction).

We were able to improve the original volume tracking pipeline by incorporating ARAP motion energy and a different notion of center affinity based on similarity of motion.

In this work, we propose a volume tracking method for time-varying surfaces of evolving topology. The main idea of the approach is that in the presence of self-contacts, the volume correspondences are more likely bijective than surface correspondences.

In this work, we propose a compression method for molecular dynamics trajectories exploiting the knowledge of molecular bonds.

We were able to improve performance of High-pass coding approach (also known as Laplacian compression) under mechanistic error metrics like MSE.