Simulation Guided Hair Dynamics Modeling from Video

The study presented a hybrid approach to reconstruct hair dynamics from multi-view video sequences captured under uncontrolled lighting. It combined image-based reconstruction techniques with physically based hair simulation, using particle-based and incompressible fluid simulations to refine hair dynamics. This method improved the motion of reconstructed hair fibers and addressed issues like occlusion or tracking failure, ensuring consistency with video inputs. The refined dynamics could also be used to create new hair animations for different styles. The method was validated with various real hair examples.