Automatic face color segmentation based rate control for low bit-rate video coding

In this paper, we propose automatic face segmentation based on skin color information and post-processing methods for low bitrate video coding applications. The proposed algorithm is used in conjunction with our proposed dynamic face based rate-control algorithm, based on the H.263 TMN8 R-D model, using an automatic enhancement weight setting (AEWS) scheme to intelligently allocate a higher number of bits to the face region. Our proposed scheme works well with a variety of video sequences, which possess different skin color characteristics and different complexity natures. Simulation results indicate that our proposed scheme encodes the video sequences with visibly higher quality on face regions with some introduced degradation to non-face regions, which are insensitive to human perception. Our proposed algorithm is also H.263 compatible, and of low-complexity, which makes the algorithm suitable for use in real time communications.