基于MPEG压缩视频的场景摘要合成算法

分析了MPEG-2的编码结构,提出在压缩域上直接提取I帧,克服了传统视频检索算法先解压视频文件再进行处理的缺陷.然后利用I帧的DC系数进行信息系统建模,通过RS的属性约简理论对I帧约简得到反映镜头主要内容的关键帧,之后对关键帧的DC系数进一步建模,根据帧的差异度得到相应镜头的分类,由此将同类镜头的I帧聚合得到场景摘要,从而获得视频的场景摘要.

The Scene Abstract Synthesis Algorithm Based on MPEG Video

MPEG-2 coding structure is analyzed and the scene Abstract synthesis algorithm directly extracting I-frame in the compressed domain is put forward,which overcomes the flaws of traditional video retrieval algorithm that extracts the video first and then process.The video scene Abstract was obtained by the following steps: construct information system modeling with the DC coefficience of I-frame;get the key frame reflecting the main content of shot through the attributing reduction theory of R S;construct the model of DC coefficience of key frames;obtain the classification of the corresponding shot according to the differences of key frames and thus the I-frame of the same classified shots are synthesized to get the scene Abstract.