Robust multiple description distributed video coding using optimized zero-padding

Anhong Wang, Yao Zhao, Huihui Bai

Abstract: Distributed video coding (DVC) arouses high interests due to its property of low-complexity encoding. This paper proposes a robust multiple description DVC (MDDVC) under the constraint of low-complexity encoding. In MDDVC, zeros are padded to each frame and the resulting big-size video is divided into multiple descriptions. Then, each description is compressed by a hybrid DVC (HDVC) codec and transmitted over different channel. When one channel does not work, the lost HDVC description is estimated by the received from other channel, which guarantees the robustness of the system; MDDVC moves the complex motion estimation totally to the decoder so it features low-complexity encoding. In the pre-processing, an optimized zero-padding is also proposed to improve the performance. Experimental results exhibit that the proposed MDDVC scheme achieves better rate-distortion performance and robustness than the referenced especially when packet-loss rate is high.