MPEG压缩域中自适应镜头实时边缘检测

Self-adaptive and Real-time Shot Boundary Detection in MPEG Compressed Domain

摘要: 提出了一种在MPEG压缩视频中利用DCT系数和运动信息快速检测镜头边缘的方法,首先利用MPEG视频中I帧DC图像间的特征值差初步判断镜头的边缘,其中的特征值差阈值能自适应调整;然后利用B帧的运动信息将镜头的边缘在图像层GOP中精确定位。实验表明,该方法能快速、准确地检测出视频中镜头的边缘。

Abstract: Video shot boundary detection is the precondition of video indexing and retrieval based on content. The Paper presents a fast algorithm for shot boundary detection using DCT coefficient and motion information in MPEG compressed video. At first, shot boundary is judged using the feature difference value between DC images of neighboring I frame; feature difference threshold may be self-adaptively regulated; And then, shot boundary is located precisely in GOP using motion information of B frame. Experiments result show that it can detect shot boundary quickly and exactly from compressed video.