Extraction of scenes containing a specific person from image sequences of a real-world scene

A method of extracting the head area from image sequences of a real-world scene for the purpose of facial discrimination processing is proposed. The method is based on tracking of the head area by characteristic point matching and edge matching. Experiments show that the method can extract the head area from scenes of a person walking freely against a general background in variable lighting, a person coming into contact with other moving objects, and when the shape of the head changes.<
>