摘要
针对传统的基于空域的视频镜头检测算法需要解压缩,计算量大、效率低的缺点,提出了一种基于压缩域的视频镜头检测方法。该方法首先根据MPEG压缩标准,从视频流中提取I帧中8个低频DCT系数,并对其进行分区加权求二次帧差,从而确定镜头变换所在的GoP;在GoP中通过计算P帧和B帧的不同类型宏块比率,来精确定位镜头边界。实验结果表明,该方法可大大减少计算时间和数据量,且具有良好的检测效果。
Aimed at the shortcoming in conventional airspace shot detection need to decompress, computationally intensive and low efficiency, a video shot detection method based on compressed domain is proposed. In this method, the 8 low-frequency DCT coefficients of I frames were extracted first from the video stream according to MPEG standard, and the frames are partitioned, weighted and computed the twice-frame-difference, so the CoP of the shot boundary is determined ; In the CoP, by calculating the different types of macro-block ratios of the P and B frames, to accurately detect shot boundary. Experiment result showed that, the amount of computation and data is reduced greatly and has a good detection results.
出处
《电视技术》
北大核心
2014年第13期7-10,共4页
Video Engineering
基金
国家"863"计划项目(2011AA01A107)
国家科技支撑计划项目(2012BAH51F02)
关键词
视频镜头检测
压缩域
MPEG
二次帧差
宏块比率
video shot detection
compressed domain
MPEG
twice-frame-difference
macro-block ratios
