Abstract: In most motion compensated video coding schemes, the compression factor deteriorates very rapidly as the signal to noise ratio (SNR) of the decoded data stream is increased. In this research, we propose an alternative approach to block based motion compensation in which both a dense motion vector field and the compensation residual are coded at different resolutions. The new approach offers several advantages over some of the well known compression techniques such as the MPEG. First, since the signal is coded at different resolutions, a low resolution signal can be extracted from the data stream without decoding the fine resolution signal. This scalable feature offers real time decoding by a wide range of different receivers with limited computational capabilities and fast forward preview at a full frame rate. Second, for moderately low SNR range, we obtain much better compression than is possible with the conventional techniques.!6
展开▼
