Abstract
This paper presents a subject of decorating human face in video image by estimation for face posture with particle filter. First, in order to estimate the face posture, we have proposed a method based on state space model with simplified two–demensions face and the positions of two eyes. Real-time estimation of face posture can be realized by particle filter algorithm .With the result of estimation of face posture, face decoration process can be done. We mainly have three achievements of face decoration: enlargement of eyes, porcelain skin and small face. For the process about enlargement of eyes, it means to calculate areas of the eyes from the positions of the pupils, and then enlarge them. Members of our laboratory were taken as the objects, video images were collected with a web camera into different conditions of face: (1)still;(2)turning left or right;(3)shifting parallelly;(4)moving back and forth;(5)swaying. From the results of experiments, face decoration process on conditions (1),(2)and(3)were realized in good performance, but conditions (4),(5)were difficult to achieve the result we want so that the state space model for estimation of face posture needs to be improved.
