Video-Based Modeling of Dynamic Hair

We present a method for creating realistic hair animation models based on videos of real hair in motion. We use an array of synchronized cameras to capture dynamic hairstyles. The hair structure and motion is modeled using an algorithm that grows each hair in segments from the root to the tip. Our algorithm generates hairs that are consistent with hair orientations observed in the input images and also enforces temporal smoothness constraints on the shape of the hair from frame to frame. Examples of synthesized hair models show the effectiveness of the method.