多视点分布式视频编码的研究

Recent advances in multi-view distributed video codecs

随着新型多媒体传感器网络的出现,现有编码方法不能有效适应其庞大的数据量.因此针对多媒体传感器网络能量、资源受限的问题,该文研究能提供低复杂度、低能耗、低成本编码器的分布式视频编码技术.文章总结了多视点分布式视频编码各个环节的最新研究进展,并针对分布式视频编码的核心内容边信息生成,提出了基于多视图几何的空间边信息生成算法,又在此基础之上应用线性组合模型的融合算法使边信息质量进一步提高.仿真结果表明,空间边信息生成算法可以使率失真性能得到2 dB左右的提高,而线性组合模型的融合方法可以使增益再提高0.3 dB左右.

Recently novel multimedia wireless sensor networks（MWSN） have appeared which have distinctive characteristics that include large amounts of image or video data.This makes conventional codecs unsuitable for an MWSN.To solve the problem,given an MWSN＇s energy and bandwidth restrictions,a distributed video codec which has low-complexity,low-power consumption,and low-cost video encoding was proposed.This paper reviews the advantages of multi-view distributed video coding and presents a new method for spatial side information generation based on multi-view geometry for use in an MWSN.Then an adaptive linear combination fusion algorithm was used to get better side information.Experiments showed that the spatial side information generation method effectively improved the quality of side information.The basic rate-distortion gain was about 2 dB and the fusion algorithm provided an additional improvement of about 0.3 dB.