Hair is a prominent feature of a character look, and hair modeling is necessary for a character animation. This chapter proposes an image-based approach to model hair geometry from images at multiple viewpoints. A hair fiber is unfortunately too fine to be discernable at the typical image resolution if the entire hair is covered in an image. But a hair fiber has a simple generic shape, which is a space curve approximated by a poly-line. Therefore, a hair fiber is reconstructed by synthesizing, segment by segment, which follows local image orientations. The hair volume is also estimated from the visual-hull-like rough reconstruction from the multiple views. We demonstrate the method on varieties of hair styles.
Unable to display preview. Download preview PDF.