摘要
提出了一种基于累积直方图的视频镜头边界检测方法,以累积直方图来代表视频帧图像的特征,其帧差充分反映了视频帧图像间的差异性;结合滑动窗的局部阈值分割处理,获得镜头边界检测对物体/摄像机的运动和光线变化的不敏感性。实验结果表明,本方法在镜头突变边界检测中达到95.97%的查全率和96.75%的查准率。
A video shot boundary detection method based on cumulative histogram is proposed, in which cumulative histogram is used to represent the character of video frames. The frame - to - frame differences of cumulative histogram fully represent the differences between video frames. Then a local threshold is selected in a sliding window for detecting shot boundary. This method is not sensitive to the movement of object/camera or the change of illumination. Experimental results show that the new method has reached the recall ratio of 95.97 percent and the precision ratio of 96.75 percent while detecting abrupt shot boundary.
出处
《电讯技术》
2008年第3期65-69,共5页
Telecommunication Engineering
关键词
数字信号处理
视频检索
镜头边界检测
突变边界
累积直方图
digital signal processing
video retrieval
shot boundary detection
abrupt boundary
cumulative histogram
