A dictionary-based HD video compressing scheme for videoconference

Currently, HD videoconference becomes more and more prevalent. The bit rate of HD videos is large, so effective compressing scheme is very important. A static or gradually changed background in a global manner is an important characteristic for videoconference. H.264/AVC, which is the state-of-the-art codec, performs well, but it is not efficient enough for videoconference since it cannot fully exploit the characteristic of videoconference. In this paper, we first train a background dictionary based on a few observed frames. And then, decompose every frame into the linear combination of dictionary and residua encoded by H.264/AVC, which dramatically removes the temporal redundancy. The experimental results on videoconferencing videos show that our method significantly reduces the bit rate of the videos with the same PSNR compared to H.264/AVC.