异种帧的相关性对MPEG-2视频源建模的影响

MPEG编码标准通常使用I、P、B三种帧类型。由于P，B帧利用基于I帧的运动补偿技术进行编码，因此同一GOP内不同类型帧之间存在较强的相关性。考虑场景并用二阶自回归模型为I帧建模：（1）对P，B帧采用相地于I帧的一阶合成回归模型建模；（2）分别对P，B帧采用二阶自回归模型建模。实验结果表明，考虑同一GOP内不同帧之间的相关性的模型，比较准确地反映了视频源的统计特征与排队性能。

Influence of Correlation of Different Kinds of Pictures on the Modeling of MPEG-2 Video Soures

I,P,B frames are usually used by the MPEG coding standards. Due to the P,B frames moving compensation based on I frame, strong correlations among different kinds of frames in the same group of pictures GOP are existed. With consideration of scene distribution and modeling I frame with second order auto-regressive model,(1) P,B frames are modeled by first order compound regressive model based on I frames, and (2) P,B frames are modeled by second order auto-regressive model respectively. The experiment results indicate that the model which takes account the correlation among different type pictures in the same group of pictures accurately reflects the characterization and queue performance of video sources.