We present a new image deformation method for mapping Peking opera facial makeup onto a human face with time varying expressions in a video sequence as shown in Fig. 1. In the mapping, the organs of the facial makeup, such like eyes, mouth, are aligned with that of the human face tightly. At the same time, the shape of artistic patterns of the facial makeup is well preserved. We use a video and a facial makeup image in vector format as input, first we define a set of points on the boundary of organs of the facial makeup as the feature points, as well as their correspondences on the human face in the first frame of the video. We track the set of points on the human face using active appearance models (AAM) algorithm [11] to get their corresponding points on each video frame. For each video frame, with the feature points as the control points, we use mov-ing least squares method to map the facial makeup onto the hu-man face. Compared with the other methods, the organs in the facial makeup and the human face are more tightly aligned and the shape of the patterns on the facial makeup is better pre-served in our method.
展开▼