论文部分内容阅读
镜头是视频分析和检索的基本单元,根据两个相续镜头处视频内容的变化情况,可以把镜头边界分为突变和渐变两种。镜头探测的方法很多,但大多都是针对某一类镜头准确性很高。文中,介绍了一种有效的镜头探测方法,此方法综合利用双直方图比较法和MPEG视频流中B帧宏块的信息来探测镜头边界,对不同类型的镜头,都取得了很高的准确性。通过对十几段不同类型MPEG_I视频节目进行试验,效果相当理想。
The lens is the basic unit of video analysis and retrieval. According to the variation of the video content at two successive shots, the shot boundary can be divided into two types: abrupt change and gradual change. Many ways to detect the camera, but most of them are for a certain type of camera high accuracy. In this paper, an effective shot detection method is introduced. This method detects the shot boundary by using the method of double histogram comparison and the information of B frame macroblock in MPEG video stream, and obtains high accuracy for different types of shots Sex. Through a dozen different types of MPEG_I video programs for testing, the effect is quite satisfactory.