分布式视频编码中相关噪声模型优化算法

在分布式视频编码中,准确的相关噪声模型(correlaiton noise model,CNM)是影响系统的关键因素。通过研究变换域中WZ帧与相应边信息的残差系数的分布关系,发现实际的残差分布与拉普拉斯分布模型存在一定的偏差。通过分析相关噪声模型的特性,提出相关改进算法。即根据残差系数的大小与实际概率置信度的关系,采用不同的分布参数,使得改进后的拉普拉分布模型更接近实际的残差分布。实验结果表明,该优化算法有效提高了系统的率失真性能,不仅降低了系统解码的码率,同时提升了最小均方误差(minimun mean square error,MMSE)重建的质量。

The correlation noise model optimization algorithm of distributed video coding

Accurate correlation noise modeling(CNM)is a key factor affecting the performance of distributed video coding. In this paper, the residual coefficients distribution between the WZ frame and its corresponding side information in the transform domain is researched. Based on the research, it is found that the actual residual distribution is not consistent with Laplacian distribution model. By analyzing the characteristics of the correlation noise modeling, the related optimization algorithm is put forward. In other word according to the relationship between the residual coefficients and the real confidence probability, using different distribution parameters, which make the improved Laplacian distribution model can more accurarely discibe the real residual coefficient distribution. The experimental results show that the optimization algorithm effectively improves the Rate-Distortion (RD) performance, not only reduces the system decoding rate,while improves the minimum mean square error(MMSE) reconstruction quality.