基于光流方程的用于镜头切变检测的特征

Optical flow feature based on the optical flow equation for video cut detection

为解决传统的运动图像镜头切变检测算法对背景噪声及镜头运动敏感的问题 ,通过分析镜头发生切变时图像中物体的运动规律 ,提出了一种基于光流方程的新的特征量——光流特征量。对运动图像序列中每一帧图像 ,计算其光流特征量 ,得到的特征量序列在镜头没有切变时保持短时平稳和较小取值 ,而在镜头发生切变时取值有显著增大。理论分析与实验结果证明 ,对该特征量运用加窗归一化的统计信号处理方法 ,可以有效地检测镜头切变 ,并且对摄像机摇动和变焦以及背景噪声不敏感 。

The optical flow feature ( F OF ) was derived from the optical flow equation and used in a statistical signal processing model to detect video cuts. Simulation results demonstrate that the proposed technique can detect cuts in the video sequence and is insensitive to the presence of background noise or camera motion, including zooming and panning.