视频编码中的一种率失真模型研究与实现

Study and Implementation on a Rate-Distortion Model for Video Coding

使用基于块的预测编码产生的视频流对包交换网络及无线网络中的传输错误异常敏感。如果能够在编码时准确估计出失真度,从而在率失真(R-D)优化的条件下选取最优的编码模式,就可以保证最佳的解码效果。研究了一种端到端的通用率失真模型,按照视频包在网络信道上传输的实际情况,通过对可能产生失真的几部分,包括编码器量化失真、网络传输失真及解码器掩盖失真,分别进行计算并应用于最新的视频编码标准H.264之中。实验表明,通用率失真模型能够较好的估算出从编码端到解码端的失真度,得到较好的编码效果。

The video packet encoded by block -based prediction encoding standard is very resistless to transmission over the packet -switched network and Internet. Optimal decoding result could be obtained if the distortion can be precisely predicted in the encoder and thus the best encoding mode can be chosen to encode the original video picture under the rate - distortion （ R - D ） optimization framework. A general end to end R - D model is studied and applied in the up to date video coding standard H. 264. The quantification distortion by encoder, the distortion by network transmission, and the concealment distortion by decoder are estimated respectively in the R - D model. Experimental results show that the distortion could be predicted pretty well.