Data-Driven Hair Modeling from a Single Image

Abstract

Hair is one of the most distinctive part of a human body, which is essential for the digitization of compelling virtual avatars. We present a data-driven approach for generating complete and complex 3D hairstyles from a single-view portrait. We construct a hairstyle database which contains more than 2000 hair models. Given a target hairstyle portrait image, we first draw a few strokes as guidance. We then search multiple best matching examples from the database based on our enhanced matching algorithm. Finally, we combine them consistently into a single hairstyle. The generated hairstyles are visually comparable to original portrait images. The reconstructed 3D hair models can be used for many applications, such as hair editing, and dynamic hair simulation.