3D face modeling with a consumer depth camera

Modeling the 3D geometry of the face is an important research topic in computer graphics and computer vision. The applications for face models include computer animation and facial analysis. Over the past few decades, several face modeling techniques have been developed using both active and passive sensors. Unlike passive sensors, which reconstruct the geometry of the face using reflected ambient light, active sensors measure the geometry of the face by emitting an external light source onto the surface of the face. As a result, techniques that utilize active sensors often produce higher quality face models, but the active sensors are typically expensive. Recently, low-cost consumer depth cameras have become widely available due to the success of Microsoft's Kinect camera. The Kinect is an active sensor that provides depth images at video rate; however, the images are often noisy and missing measurements. In this thesis, we present a method for modeling the geometry of the face using a consumer depth camera. To construct a high quality model, we combine the surface measurements from multiple depth images. By registering and integrating a sequence of depth images, we are able to recover the entire surface of the face and reduce noise. We demonstrate that 3D face models built using our proposed method are comparable to models generated using an expensive, high-resolution 3D scanner