Low-complexity low-delay distributed video coding

Most state-of-the-art distributed video coding schemes are based on Stanford Wyner-Ziv video coding architecture which takes advantages of efficient channel codes such as LDPC or Turbo codes. The drawbacks that prevent it from practical applications are high decoding complexity and especially latency, since it needs to decode channel codes iteratively and require feedback information many times when decoding a WZ frame. In this paper, we proposed a novel distributed video coding scheme which needs no iterative decoding and only requires feedback information once per WZ frame. It uses a block classification coding strategy combined with a residual computing method based on modular arithmetic. Our scheme replaces complex channel coding with simple entropy coding, which greatly reduce decoding complexity and latency. Experimental results show that the proposed scheme can improve the rate-distortion performance over some state-of-the-art distributed video coding schemes for low and medium motion video sequences, while reducing decoding complexity to 10%~1%.