Simulation Guided Hair Dynamics Modeling from Video

The paper presented a hybrid approach to reconstruct hair dynamics from multi-view video sequences captured under uncontrolled lighting conditions. The method combined image-based reconstruction techniques with physically based hair simulation to refine hair dynamics. An initially reconstructed sequence of hair fiber models was improved using particle-based simulation and incompressible fluid simulation, allowing for the enhancement of hair fiber motions and the completion of missing fibers due to occlusion or tracking failure. The refined space-time hair dynamics were consistent with video inputs and could be used to generate novel hair animations of different hairstyles. The method was validated through various real hair examples.