摘要
自从MPEG-4和MPEG-7标准公布以来,基于内容的视频编码与基于对象形状的检索成为视频领域新的发展趋势,而准确地从背景中分割出视频对象是实现上述功能的前提条件。对视频运动对象分割算法进行了研究,该算法结合对称差分和自适应阈值选取,获得变化检测模板,并最终分割出视频运动对象。实验结果表明该算法对运动对象敏感,能实时准确地分割出视频运动对象。但对象的关节运动以及背景的全局运动都将导致分割精度的降低,这也是今后研究的重点。
Due to the emergence of MPEG-4 and MEPG-7,content-based video encoding and object-based shape retrieval has become a new trend in video domain. But first of all, we must accurately segment the video object in video frame. The moving object segmentation in video sequence has been researched in the article. The algorithm combines symmetrical DFD and adaptive thresholding approach to get CMD,and finally segment the moving object . The experiment indicates this algorithm is sensitive to the moving object, and segment the moving object real-timely and accurately. But the joints motion of object and global motion of background will reduce the accuracy of segmentation, this is the point of our further research.
出处
《微电子学与计算机》
CSCD
北大核心
2007年第12期40-43,共4页
Microelectronics & Computer
基金
国家自然科学基金项目(60673092)
教育部科研重点项目(205059)
江苏省高技术研究计划项目(BG2005019)
关键词
视频对象分割
对称差分
自适应阈值选取
变化检测模板
video object segmentation
symmetrical DFD
adaptive threshold selection
change detection mask
