Abstract: According to the correlation noise modeling between side information and source in the Distributed Video Coding(DVC), a novel refinement method based on Multi-Hypothesis Motion-Compensated Prediction(MHMCP) is proposed. It generates original Side Information(SI) by traditional methods, uses bi-directional motion estimation to generate motion compensated blocks, and linearly combines them to generate new SI. Experimental results show that the proposed strategy with the advantage of low algorithm complexity can significantly increase the accuracy of SI, thereby effectively improves the RD performance of DVC.

Key words: Distributed Video Coding(DVC), Side Information(SI), motion estimation, Multi-Hypothesis Motion-Compensated Prediction (MHMCP), noise model

摘要： 根据分布式系统中边信息和原信息之间的噪声模型，提出一种基于多假设运动补偿的边信息改进方法。使用传统方法生成原始的边信息，对其进行双向运动估计产生补偿块，将补偿块线性组合成新的边信息。实验结果表明，该算法具有复杂度较低的优点，能提高边信息的质量，从而有效地改善分布式视频压缩的率失真性能。

关键词: 分布式视频编码, 边信息, 运动估计, 多假设运动补偿预测, 噪声模型

CLC Number: 

• TN919.81