Real-Time Face Tracking and Replacement

In this paper we present an application on real-time face tracking and replacement. We implement this application through three sections: detection, tracking and replacement. For detection, we use haar features based cascade classifiers to detect the face, eyes and nose on both target and source images. Then face tracking is implemented by two distinct methods: CAMShift and Optical Flow. In cases users choose replacement, we cover the target face using a cartoon mask or blend it with a star face using Laplacian Pyramid Blending. To reduce the difference of skin colors between faces, we modify the Laplacian Pyramid Blending to better blend the target face with the source. The experiment shows that our methods lead to satisfactory tracking and replacement result, and our application works efficiently on mobile devices.