A fixed rate multiple description three-dimensional subband video coder

Multiple description coding refers to encoding an information source into multiple bit streams in a way to be able to trade off between channel protection and decoding quality. In this paper we propose a two-description three-dimensional subband video coder. We use over-sampling to produce redundancy for the visually important subbands, and then split the subbands samples between the two descriptions. In the quantization stage, we use trellis scalar quantizer for the two descriptions of the subbands. In order to have a feedback on channel effect on video bit-stream, we add some hidden data to the texture area of video scene in the high frequency subbands. At the receiver, we check the hidden bits and select the description that has been less corrupted. We experiment the system performance using several standard video sequences and report the PSNR of reconstructed video sequences in different situations.