Automatic Facial Feature Extraction in Model-Based Coding

Model-based coding is a new image sequence compression technique for very low bit rate coding. Most researchers have paid more attention to facial image analysis and synthesis because facial images are very important in multimedia communication applications, such as video phone, video conferencing, remote learning etc. This coding method represents the image content in a structural way. This is the reason why modelbased coding can get a higher compression ratio (1-10 kb/s) than can the conventional coding methods. In order to encode image signals efficiently, it is necessary to create a suitable generic model and adapt it to the actual object accurately. In this paper, we propose a scheme, based on the integral-projection algorithm, which adapts a face model to an actual face automatically. First, the image is preprocessed by means of edge detection. From the edge mapped image, we use the feature of the increasing sudden edge density to indicate the rough vertical positions of eyes and mouth. According to the local threshold value in each feature area, all of the control points about the eyes and mouth are found. Finally, we use these control points to adapt the model to an actual face.