首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于场景参考帧的预测帧刷新
引用本文:艾育华,叶梧,冯穗力,胡兵,谢明.基于场景参考帧的预测帧刷新[J].华南理工大学学报(自然科学版),2007,35(5):10-14.
作者姓名:艾育华  叶梧  冯穗力  胡兵  谢明
作者单位:华南理工大学,电子与信息学院,广东,广州,510640
基金项目:国家自然科学基金资助项目(69972015),广东省自然科学基金资助项目(31391)
摘    要:为了阻止无线网络视频传输中由于丢包和误码而引发的差错扩散,提出了一种基于场景参考帧的预测帧刷新的抗差错策略.通过对视频序列的分割,对场景内需要同步刷新的视频帧进行基于场景参考帧的预测编码,获得相应预测帧,在编解码端同步场景标志和刷新标志,从而实现预测帧刷新.试验结果表明,与传统的I帧刷新策略相比,在需要高I帧刷新率的不可靠无线网络中,文中策略能够在保持相似重建视频质量的前提下,使帧比特率显著下降.

关 键 词:抗差错  预测帧刷新  场景参考帧  视频传输  无线网络
文章编号:1000-565X(2007)05-0010-05
修稿时间:2006-03-24

Predictive Frame Refreshment Based on Scene-Context Reference Frame
Ai Yu-hua,Ye Wu,Feng Sui-li,Hu Bing,Xie Ming.Predictive Frame Refreshment Based on Scene-Context Reference Frame[J].Journal of South China University of Technology(Natural Science Edition),2007,35(5):10-14.
Authors:Ai Yu-hua  Ye Wu  Feng Sui-li  Hu Bing  Xie Ming
Institution:Sehool of Eleetronie and Information Engineering, South China Univ. of Teeh. , Guangzhou 510640, Guangdong, China
Abstract:In order to avoid the error propagation due to packet loss and bit error during the video transmission in wireless network,a novel error-resilient scheme with predictive frame refreshment is proposed based on scene-context reference frame.By segmenting the video sequences,the corresponding predictive frames locating at the original frames and needing to be intraframe refreshed are predictably coded based on the scene-context reference frame within the same scene scope.The predictive frame refreshment is then implemented by means of the scene-flag and refresh-signal synchronization between the encoder and the decoder.As compared with the traditional intra-frame refreshment method,the proposed scheme can obviously reduce the frame bit rate with a similar reconstructed video quality when the video transmits over an unreliable wireless network that requires high intra-frame refreshment rate.
Keywords:error resilience  predictive frame refreshment  scene-context reference frame  video transmission  wireless network
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号