Multiple description coding using multiple reference frame for robust video transmission

Multiple description coding is an error-resilient coding method that encodes a video sequence into several bitstreams, called descriptions. The main problem of multiple description coding is the mismatch of reference frames in encoder and decoder, when the descriptions are lost during transmission. The mismatch is also referred to as drift and can be alleviated efficiently by transmitting side information along with each description. We introduce a new generation scheme for side information, which is based on multiple reference frames. For each encoding frame, two previously encoded frames are used as the reference frame´s I motion prediction. The prediction results of two reference frames are encoded as descriptions and side information respectively. When transmission errors occur in either reference frame, the decoder can use the correctly received reference frame in decoding. Simulation results prove that the proposed method can improve video quality obviously along different macroblock loss rates.