A new approach to wire-frame tracking for semantic model-based moving image coding

摘要

Automatic wire-frame fitting and automatic wire-frame tracking are the two most important and most difficult issues associated with semantic-based moving image coding. A novel approach to high-speed tracking of important facial features is presented as a part of a complete fitting-tracking system we have developed. The method allows real-time processing of head-and-shoulders sequences using software tools only. The algorithm is based on eigenvalue decomposition of the sub-images extracted from subsequent frames of the video sequence. Since each facial feature (the left eye, the right eye, the nose and the lips) is tracked separately, the algorithm can be easily adapted for a parellel machine. The algorithm was tested on numerous widely used head-and-shoulders video sequences containing speaker's head pan, rotation and zoom with remarkably good results. The experiments we have carried out prove that it is possible to maintain tracking even when the facial features are partially occluded.