高效率的多视点视频编码预测结构*

预测结构是多视点视频编码(Multi-View Video Coding,MVC)研究的主要内容之一。MVC目前采用HHI(Heinrich-Hertz-Institute)提出的分层次B帧预测结构(HBP),比联播预测结构获得了更好的压缩效率。分析了多种预测结构,并针对平行摄像机采集的多视点视频序列,提出了一种新的预测结构AS_EIPP,该结构充分利用相邻视点间的相关性和多参考帧模式,进一步提高了压缩效率。在多视点视频软件测试平台JMVC8.3上进行验证,实验结果表明:新的预测结构在保证重建视频质量基本不变的前提下,压缩效率比HBP预测结构提高了1%～4%。

Efficient prediction structure of multiview video coding

Prediction structure is one of the main content of study of MVC. MVC currently takes Hierarchical B frame prediction structure presented by HHI, which has better compression efficiency compared to simulcast prediction structure. This paper reviews various prediction structures, and puts forward a new prediction structure called AS_EIPP for the sequences of multi-view video obtained by parallel cameras, which take full advantage of the correlation of adjacent views and modes of multiple reference frames to further improve the compression efficiency. Measures are done on software testing platform for multi-view video called JMVC8.3.The experimental results show that the compression efficiency of the new prediction structure has increased from one percent to four percent compared to the HBP prediction structure adopted by JMVC under the premise of unchanged reconstructed video quality.