基于等效MSE的空间-时间可伸缩码流提取方法

An extraction algorithm based on equivalent MSE method for temporal-spatial SVC

可伸缩视频编码(scalable video coding,SVC)可以有效地将视频的码流通过不同网络传输给不同的用户.通过对一个帧组(GOP)中所有宏块的尺寸和运动情况进行分析,提出了等效均方误差(Eq-MSE)方法,以此计算视频不抖动的最小帧率,再根据网络的带宽,确定提取子流的时间和空间增强层的层数.这种基于视频内容的子流提取方法,有效地弥补了当前视频编码不考虑视频内容这一不足,在一定带宽下,使视频的解码质量达到最优.对比于不考虑视频内容的方法,提出的自适应子流提取方法使重建的视频峰值信噪比在一定比特率下显著改善.

Scalable video coding（SVC）is an efficient and effective solution to stream video over heterogeneous networks to different clients.An equivalent MSE（Eq-MSE）method is proposed by analyzing the motion and texture in a group of picture（GOP）to calculate the minimum frame rate of non-shaking video.According to the content of a video sequence which is temporal or spatial,it determines how many temporal and spatial enhancement layers should be extracted.The method has revised the drawback which doesn′t consider the video content effectively and improved video decoding performance at some band width.Experimental results show that the adaptive extraction method for sub-stream improves the PSNR of reconstructed video significantly under a certain bit rate.