Realtime face object extraction algorithm for video phone

This paper devises a real-time face object extraction algorithm for moving pictures. Considering that most of the existing object extraction algorithms are of too intensive computational complexity to enable the real-time execution to be applied to mobile video phone run on an embedded processor, this algorithm intends not only to reduce the computational complexity by employing modified snakes but also to improve the precision of extracting pixels by an efficient estimation mechanism of the skin color portion within a face object. Simulation results demonstrate that by this algorithm QCIIF 15fps can be processed on a StrongARM processor in real-time.