一种从MPEG视频流中快速分割镜头的算法

A Fast Algorithm of Scene Change Detection for the MPEG Compressed Video Stream

镜头分割算法是实现视频检索的关键,本文提出一种针对MPEG视频流的快速分割镜头的算法,介绍了该算法的基本原理与实现过程。该算法利用压缩视频中的可用信息,通过计算帧的内编码宏块数量的比例作为是否需要进行镜头分割判别的标识,进而计算MPEG视频流B帧各宏块编码类型来检测发生镜头分割的帧位置,并进行了相关的实验。实验结果表明,该算法能快速有效地检测镜头分割,且计算量较小。

The scene change detection algorithm is the key to video content-based retrieval. In this paper we propose an algorithm that detects scene changes from the MPEG compressed video bit stream, and explain the principle and how to implement the algorithm. This algorithm is based on the comparison of the majority of intra-macrro-blocks and the macro-block type of B-frames from the bit stream. By partly decoding on the bit stream, computational complexity is reduced and experimental results show that the presented algorithm can provide satisfactory shot segmentation results.