Layered multiple description video coding using dual-tree discrete wavelet transform and H.264/AVC

A novel hybrid layered multiple description video coding (HLMDVC) algorithm is proposed to provide a robust and flexible video transmission via unreliable networks, like Internet or the wireless network. First, the input video sequence is encoded by a standard H.264/AVC encoder in low bit rate to form the base layer, which is then duplicated to each description. Second, the 3D dual-tree discrete wavelet transform is performed on the difference between the reconstructed base layer and the input sequence to produce four wavelet coefficient trees. Then the noise shaping is used to reduce the redundancy of these coefficient trees. After that, the four sparse wavelet coefficient trees are separately encoded by 3D-SPIHT encoders. The resulted four bitstreams are partitioned into two groups, separately forming enhancement layers of two descriptions. The redundancy of the proposed multiple description video coding scheme can be easily adjusted by changing the bitrate of the base layer to fit for different transmission channels. The directional selectivity feature of 3D dual-tree discrete wavelet transform frees the enhancement layer coding from the time consuming motion estimation, such that no mismatch happens and the coding efficiency is improved. Simulation results have shown that the quality of the reconstructed video of the proposed HLMDVC algorithm is superior to that by the state-of-the-art multiple description video coding methods.