Face Detection and Identification in Video-Footages from Key Face Images

Previous studies on face detection in video footages show that segmenting faces accurately and reliably is often hard to succeed, leading to interactive manipulation. This paper presents a simple and practical face detecting and identifying system in video footages from input key face images with focus on segmentation. Each time the scene change is detected the system tries to find a similar face on the beginning frame to each of the key faces. We present an improved color-based segmentation that uses RGB and HSV systems and includes isolated but enclosed face parts, to generate more informative binary face patterns with representative color information. A few variations of the face image are synthesized from the input key face images, to more effectively detect faces in the video footage that may slightly be tilted or rotated from the frontal one. Some experimental results are given.