摘要
经过对视频序列的多帧累积,利用帧间变化信息可求得帧差之和,并用数学形态学方法处理之后,即可提取出对象边界。由此,提出了一种累积帧差的视频对象分割算法。实验结果表明,该算法能对低比特率视频序列中运动较为缓慢、背景相对简单的对象进行有效的分割,而且所用时间较短,算法简单、高效。
Through multi-frame accumulation for video sequence, the sum of frame differences can be obtained with interframe variation information. After the obtained results being processed with mathematical morphology, the object boundary can be extracted. A video segmentation algorithm based on accumulated frame differences is proposed from this point. The experimental results show that the object with slower motion velocity and relative simple background in a low bit rate video sequence can be effectively segmented by this algorithm.
出处
《光电工程》
CAS
CSCD
北大核心
2004年第7期69-72,共4页
Opto-Electronic Engineering
关键词
图像分割
帧差图像
数学形态学
Image segmentation
Frame difference image
Mathematical morphology
