Method and apparatus for accomplishing multiple description coding for video

Abstract

A method and apparatus for utilizing temporal prediction and motion compensated prediction to accomplish multiple description video coding is disclosed. An encoder receives a sequence of video frames and divides each frame into non-overlapping macroblocks. Each macroblock is then encoded using either an intraframe mode (I-mode) or a prediction mode (P-mode) technique. Both the I-mode and the P-mode encoding techniques produce an output for each of n channels used to transmit the encoded video data to a decoder. The P-mode technique generates at least n+1 prediction error signals for each macroblock. One of the at least n+1 P-mode prediction error signals is encoded such that it may be utilized to reconstruct the original sequence of video frames regardless of the number of channels received by the decoder. A component of the one of the at least n+1 P-mode prediction error signals is sent on each of the n channels. Each of the remaining at least n+1 P-mode prediction error signals is sent on a separate one of the n channels (along with the above mentioned component). These remaining at least n+1 P-mode prediction error signals are encoded such that, when combined with the component of the one P-mode prediction error signal which was sent on the same channel, a reasonably good reconstruction of the original sequence of video frames may be obtained if the number of received channels is between 1 and n−1.