摘要
提出了一个直接利用MPEG视频中的压缩参数信息,获取镜头运动特征的算法.算法中利用的压缩参数主要有帧类型、宏块类型、DC系数、运动矢量.与已有的算法相比,此算法具有计算成本低,识别率高的特点,能有效地区分镜头运动和溶暗、划出等镜头渐变,具有鲁棒性.
We present an algorithm that acquires shot motion features using compressed parameters by MPEG provided. Compressed parameters include type of frame, type of macroblock, DC coefficients, and motion vectors. Compared with the exist algorithms, the algorithm has lower cost of computing and higher recognized ratio, and it can distinguish motion of shot and gradual transition efficiently, such as dissolve and wipe. It is robust against noise and moving objects.
出处
《武汉大学学报(理学版)》
CAS
CSCD
北大核心
2004年第3期330-334,共5页
Journal of Wuhan University:Natural Science Edition
基金
国家自然科学基金资助项目(60173045
F020204)