一种在MPEG压缩域中检测镜头扫换的方法

当前对镜头切变已能够提供有效的检测方法,但渐变的检测一直是一个难题,也是当前视频分割新的研究热点.本文提出一种在MPEG压缩域中检测扫换(Wipe)的方法,它通过分析B帧和P帧中宏块编码类型的时空分布特点来进行检测,利用哈夫变换判断变化区域连续性,最后使用DC图像的直方图分析进一步验证.它不仅能定位扫换的起止帧,而且能得到扫换方向等特性.本方法需要比较小的运算量,同时对运动和其他镜头切换有比较好的区分性.

A Wipe Transition Detection Method In MPEG Compressed Domain

Now many video segmentation algorithms can detect cuts efficiently,but the detection of gradual transitions is left as a challenge,which is becoming a new active research region.In this paper,a new wipe transition detection method in MPEG compressed domain is presented based on the spatio temporal distribution of the macroblock coding types in B frames and P frames.Hough transform is utilized to analyze the continuity of change regions.At last,the histogram analysis of DC image is used to reconfirm the detection.This method can get not only the beginning and end frame number of the wipe transition,but also the transition patterns,such as the wipe direction.It requires less calculations and can distinguish wipes from other transition types and motions in some cases.