本文提出了一种基于H.264/AVC压缩域的GOP(Group of Pictures)级视频场景转换检测算法.该算法利用H.264/AVC基本档次码流中的帧内预测模式、运动矢量和宏块编码类型等可用信息,提出了基于子块的色度模式差异、累积运动值和累积帧内宏块数等三个判决准则,然后综合利用这三个判决准则,提出了一种GOP级的视频场景转换检测算法.实验结果表明,与现有的一个GOP级场景检测算法对比,本文提出的COP级视频场景转换检测算法可以获得更好的检测性能.
A GOP level video scene change detection algorithm in H.264/AVC baseline profile compression domain is proposed. The algorithm is based on intra prediction modes, motion vectors and macroblock coding types which can be extracted from H.264/AVC baseline bitstream. Three new criteria,the difference of chroma prediction modes based on sub-block,the accumulative motion amount and the accumulative intra coding macroblock amount, are proposed. Experimental results show that compared with another typical GOP level method, the proposed algorithm can achieve better performance.