SNR-scalable extension of H.264/AVC

Abstract

We present an SNR-scalable extension of the H.264/AVC video coding standard. To achieve an efficient SNR-scalable bit-stream representation of a video sequence, the temporal dependencies between pictures are exploited by using an open-loop subband approach. The related temporal analysis-synthesis filter bank structure is generalized to facilitate an adaptive block-based choice between the motion-compensated lifting representations of the Haar filter (uni-directional prediction) and the 5/3 filter (bi-directional prediction), both coupled with multiple-reference frame capabilities. As a remarkable feature of our approach, most components of H.264/AVC are used as specified in the standard, while only a few have been adjusted to the motion-compensated temporal filtering structure. Our proposed SNR-scalable extension was tested for a set of CIF sequences, and the results indicate that a coding efficiency comparable to that of the state-of-the-art H.264/AVC standard can be achieved.