细粒度扩展视频均等质量流化算法

Algorithms for Equal-Quality Fine Granularity Scalable Video Streaming

细粒度扩展(fine-granularity-scalability,简称FGS)编码具有很强的灵活性和较好的视频流化性能,故已被MPEG-4和H.26L等标准所采用.FGS编码的一个突出特点是可以随意裁减以适应网络带宽的变化.但是,简单的裁减方法容易造成连续图像质量抖动过大,而用户通常希望流化视频的图像质量尽可能地平滑.在相关研究的基础上,针对非实时流化应用,分别讨论了在无损和有损条件下进行FGS编码均等质量流化的算法.该算法基于FGS的分段线性率失真模型和滑动窗口,在无损条件下,使用二分法,在当前窗口中的所有帧之间进行速率分配,以实现均等质量流化;在有损条件下,通过自适应的启发式算法,并结合前向纠错(forwarderrorcorrection,简称FEC)技术来达到同样的目的.实验结果表明在两种情况下,该算法均可以获得较好的流化效果,使流化视频的图像质量更加平滑.

Fine-Granularity-Scalability （FGS） coding can provide the flexibility and good performance for video streaming, and thus has been accepted in MPEG-4 and H.26L. It is noticed that FGS can be truncated anywhere to adapt the changes of the network bandwidth. However, the existing simple truncation methods would lead to the fluctuation of the quality of video sequence. Often, users would like to obtain a smooth quality. Based on the discussions of related works, this paper presents two equal-quality FGS video streaming algorithms with/without loss in the non-real-time streaming application scenarios. These algorithms are based on the piece-wise line R-D model and the slide window protocol. In the lossless case, the bisection method is exploited to allocate the network bandwidth among the frames in the current window so as to realize the equal-quality FGS video streaming; while in the loss case, an adaptive heuristic algorithm and the forward effort correction （FEC） technique are used for the same purpose. The experimental results demonstrate that the algorithms can have good performance in both of the two cases, thereby obtaining the smoother streaming quality of FGS video sequence.