An MPEG-4 virtual video conferencing system with robust video object segmentation
Jinhui Pan, Chuang Gu, Ming–Ting Sun · 2002
We introduce a practical MPEG-4 virtual conferencing system using a robust video object segmentation algorithm. This algorithm takes advantage of high-order statistic features of video sequences and the relation between continuous frames to get the final segmentation. The method can produce satisfying results with pixel-wise precision and is robust to the effects of noise and the changing of lighting condition.