Abstract: A novel Multiple Description video coding scheme is proposed based on Predictive Error Resilience (MD-PER). At the encoder, the possible error caused by the single-channel reconstruction can be predicted firstly, and then the necessary redundancy information is inserted for each description. In view of the compression efficiency, different coding modes are designed to compress the generated redundant information. At the decoder, the redundancy information can be applied to achieve high quality recovery of the lost video frames. Experimental results demonstrate that compared with the traditional temporal sampling method the proposed scheme achieves better rate-distortion performance.