多参考预测链的时域可扩展性和抗错性

Error resilience and temporal scalability for video transmissions using multiple reference prediction chain coding

网络的带宽抖动和数据传输错误是影响视频传输的重要原因。该文提出了一种新的多参考帧图像预测的视频编码算法,即多参考预测链,使视频具有时域可扩展性和抗错性解决网络的这两个潜在问题。该算法中P帧之间的参考预测关系构成多条链状结构,使视频具有时域可扩展性。链间的不相关性使每个链可独立解码,从而增强了码流抗错性。该算法采用R-D框架动态的管理预测关系链,从而适合网络带宽的变化而获得最佳的终端视觉效果。仿真实验表明,采用多参考预测链编码的码流时域可扩展性的应用非常可行,且表现出较高的抗错性。

Bandwidth and transmission errors are two common problems in network video transmissions. This paper presents a multiple reference picture selection scheme to encode video for temporal scalability and error resilience. Temporal scalability is obtained with prediction chains constructed to give prediction dependency between P-frames. The inter-chain independency also endows the video with error recovery ability. The prediction chains for a group of frames are dynamically managed using a rate-distortion optimization framework so that the expected end-to-end video quality is achieved for a given rate constraint. Simulation results show that the algorithm provides good temporal scalability and error resilience.