分布式多视点视频编码中的立体边信息生成  

Generation of Stereo Side Information in Distributed Multi-view Video Coding

作　　者：刘爽[1] 史萍[1] 任培明[1] 

机构地区：[1]中国传媒大学信息工程学院,北京100024

出　　处：《电视技术》2012年第19期27-31,共5页Video Engineering

基　　金：亚洲传媒研究中心重大项目(AM0501)

摘　　要：新兴的分布式多视点视频编码通过在解码端使用立体边信息可充分挖掘传统单视点的时间相关性和多视点特有的视间相关性。研究了一种多视点运动预测的立体边信息生成算法,提出了近似视差估计和编码端掩模融合算法。实验证明,解码端采用近似的视差矢量,编码端传递某些先验信息,都能有效提高立体边信息精度,降低计算复杂度。Distributed multi-view video coding, which uses side information at the decoder, can exploit traditional temporal correlation within the single view video as well as the inter-view correlation between muhi-view videos. One of the stereo side information generation technologies, namely multi- view motion estimation, is studied in this paper. The approximate disparity estimation and encoder mask fusion algorithms are further proposed. Experiments show that, by using the approximation disparity vectors or transporting some a-priori information from encoder, the accuracy of stereo side informa- tion can be improved while the computational complexity is reduced.

关 键 词：多视点视频编码 分布式多视点视频编码 立体边信息 视间预测 

分 类 号：TN919.8[电子电信—通信与信息系统]