摘要
在兼顾运动图像分割效果和实时性的基础上,针对视频序列,仅利用其时域信息,提出了一种简单有效的运动前景分割算法。首先对图像序列做帧间差分与隔帧差分,然后将两类差分结果进行累积,对累积结果采取交集聚类的方法求出运动前景轮廓。二值化处理后,扫描填充即可得到图像序列中的运动前景。实验证明:该算法兼顾分割效果和实时性的要求,是一种较好的运动前景分割算法。
Based on the rules of the segmentation result and real time requirement,this paper represents a simple moving foreground segmentation method in video sequences only using their temporal information.Firstly,we obtain the changing region by the intersection clustering of two kinds of accumulation of the consecutive and discontinuous frame-difference.Secondly,after binarization and scanning the moving region,we can get extracted moving foreground by filling moving region with original sequence.Experiment results show that the method can successfully segment the moving object from the video sequences with less time,which is a better moving foreground segmentation method.
出处
《计算机工程与应用》
CSCD
北大核心
2007年第20期229-230,239,共3页
Computer Engineering and Applications
关键词
图像分割
运动前景
累积
聚类
image segment
moving foreground
accumulation
clustering
