MPEG-4框架下的视频对象分割及其关键技术分析?  

Video object segmentation under the framework of MPEG-4 and its key techniques analysis

作　　者：杨高波[1] 张兆扬[2] 

机构地区：[1]湖南大学计算机与通信学院,湖南长沙410082 [2]上海大学新型显示器技术及应用集成教育部重点实验室,上海200072

出　　处：《通信学报》2007年第10期109-122,共14页Journal on Communications

基　　金：国家自然科学基金资助项目(60572127);湖南省自然科学基金资助项目(05JJ30113)~~

摘　　要：在简要介绍MPEG-4引入的视频对象概念及其应用背景的基础上,系统地研究了视频对象分割所涉及的关键技术。首先,按照视频是否压缩将现有算法划分为像素域和压缩域两类,详细讨论了所涉及到的初始对象分割、对象跟踪技术。其次,阐述了立体视频对象分割以及基于细胞神经网络体系结构的对象分割算法的研究进展。此外,对视频对象分割算法的客观性能评价进行了总结。最后,对本领域值得进一步研究的问题进行了预测和展望。After briefly introduction of the concept of video object under the framework of MPEG-4 and its application background, the key techniques involved in video object segmentation was systematically researched. Firstly, currently available algorithms were classified into two categories： compressed-domain based and pixel-domaln based algorithms, and then the involved initial object extraction and object tracking are discussed in detail. Secondly, the recent advances of video object segmentation in stereo video sequences and video object extraction in the cellular neural networks were also presented. Thirdly, the objective performance evaluation of different segmentation algorithms is summarized. Finally, future directions worthy of further investigation on this topic were also prospected.

关 键 词：视频对象分割 MPEG-4 压缩域 细胞神经网络 分割评价 

分 类 号：TN913.24[电子电信—通信与信息系统]