Face detection in the compressed domain

Face detection is important in many algorithms in the areas of machine object recognition and pattern recognition. The kaleidoscope of applications for face detection extends across automatic image and home video content annotation, face-image stabilisation and face recognition systems. By using information derived from colour, luminosity and frequency the face detection algorithm proposed in this paper aims to determine the location of multiple frontal and non-frontal faces in compressed MPEG-1, MPEG-2 video or JPEG image content. The described algorithm requires only low computational resources on CE devices and offers at one and the same time extremely high detection rates.