摘要
镜头边界检测是视频分析的重要一步。提出一种基于神经网络的检测算法,同时使用直方图差值和差分直方图方差,对亮度变化和物体运动具有鲁棒性。试验证明该检测算法具有较高的准确性。
Shot boundary detection is an important step in video analysis, A neural-network based detection method is proposed using both histogram difference and variance of difference histogram, which make it robust to light changes and object movement. Experiment results show that this method has a high accuracy.
出处
《计算机工程》
CAS
CSCD
北大核心
2006年第5期181-182,共2页
Computer Engineering
基金
国家自然科学基金资助项目(60273064)
关键词
镜头边界检测
差分直方图方差
直方图差值
神经网络
Shot boundary detection
Variance of difference histogram
Histogram difference
Neural network
