Facial Component Extraction by Cooperative Active Nets with Global Constraints

Ryuji Funayama, Naokazu Yokoya, Hidehiko Iwasa, Haruo Takemura

Extraction methods of face and facial components are essential and expected to be applicable to personal identification, computer graphics, advanced human interface, etc.

This paper describes a new face and facial component extraction method from a face image using cooperative active nets. An active net model which is based on energy minimization principle is used for extracting non-rigid object region and I utilize this model for extracting face and facial components. Image force of a net is defined from color distribution of a face region and external constraint force from size, shape and symmetry of facial components. Each net deforms with its own energy being minimized and the position of the nets are controlled to minimize an energy which is produced by constraints of their placement.

Our method has been experimentally shown to be robust to variations of facial size, position and rotation compared with existing methods such as template matching and projection-based method. And it is not required to be carefully initialized unlike deformable templates.

Ryuji FUNAYAMA <post@funayama.com>
inserted by FC2 system