摘要
在兼顾运动图像分割效果和实时性的原则上,针对视频会议、新闻播报等低比特率视频序列,提出了一种简单高效的运动对象分割算法。首先利用累积帧差求出图像帧的运动区域,然后对其进行二值化和形态学处理得到帧差模板,最后利用二次扫描的方法得到运动对象掩模,对其进行填充就可以提取出运动对象。实验证明得到了较好的分割效果并且在实时性的应用中具有一定的优势。
Based on the rules of the segmentation result and real time requirement,aiming at low bit rate video sequences,an effective moving object segmentation algorithm is proposed.First we obtain the moving regions by accumulating the frame-differencing.Then we obtain the frame difference mask by thresholding and mathematical morphology processing.Finally,the moving object mask is scanned by every row and column,and filling moving object mask with original values gets extracted moving object.Experimental results demonstrate that the method can get a good result and it is an appeal for real-time applications.
出处
《计算机工程与应用》
CSCD
北大核心
2006年第19期155-157,共3页
Computer Engineering and Applications
关键词
低比特率
视频对象分割
累积帧差
掩模
low bit rate,video object segmentation,accumulative frame-difference,mask
