分布式视频编码中基于滑窗的自适应相关估计方法

Sliding-window based adaptive correlation estimation method for distributed video coding

信源与边信息之间的相关建模和参数估计一直是分布式视频编码的关键环节。在分布式视频编码中,相关噪声分布是非平稳的,会随着场景序列而动态变化,如何准确地预测和追踪相关参数十分重要。为了充分发挥分布式信源编码的优势,提出一种基于滑窗的自适应相关估计方法。该方法将滑窗的思想嵌入到基于LDPCA码的Slepian-Wolf解码器中,结合边信息和码流在线估计图像信源之间的相关参数,且能够自适应地选择是否进行重估,进而优化联合比特面置信传播迭代译码。实验结果表明,以先进的分布式视频编码方案DISCOVER作为基准,使用所提方法的方案在保持较低编码复杂度的情况下提高系统的率失真性能。

The correlation modeling and parameter estimation for signal sources and side information are key processes in distributed video coding(DVC). In DVC,correlation noise distribution is unstable and changes dynamically with the scene sequence,so it is important to accurately predict and track correlation parameters. To give full play to the advantage of distributed source coding(DSC),an adaptive correlation estimation method based on sliding-window is proposed in this paper. In the method,the sliding-window idea is embedded in the Slepian-Wolf decoder based on LDPCA code. The method can adaptively select whether re-estimation should be executed in combination with correlation parameters of side information and bit stream online estimation image source,so as to optimize joint bit-plane belief propagation(JBBP)iterative decoding. The experimental results show that,taking the advanced DVC scheme DISCOVER as the criterion,the scheme using the proposed method can improve rate-distortion performance of the system while maintaining low coding complexity.