Production of facial expressions using facial feature positioning and deformation

Based on image processing technique, this paper used face images of users as virtual images to automatically produce different facial expression images. The face images in neural state can be transformed into the facial expression images. First, for the face images in neutral state, skin color segmentation and morphological noise processing were conducted to separate the possible face zone from the background. In terms of feature points (FP) and facial animation parameters (FAP) defined in MPEG-4, feature shape control points were marked for features of the facial expressions: color information of brows, eyes, and mouth. The feature control points were marked on the images during production of the facial expressions. Delaunay Triangulation, image registration and image interpolation were performed to move the feature control points to change face images through interactive interface. The image texture was changed by adjusting positions and shapes of feature texture in order to automatically produce new facial expression images.