摘要
文章在宏块层的码率控制中使用线性R-Q模型,该模型与二次R-Q模型相比可以更好地克服模型参数抖动的问题。在R-Q模型参数更新时,提出了改进的2D滑动窗口进行数据选择,相对于JVT-H017提案里的窗口更加充分地挖掘了空间相关性;同时提出了新的MAD预测模型,该模型相对于JVT模型达到了更加精确的预测效果。仿真实验结果表明,该算法与原JVT相比峰值信噪比得到了明显的提高,比特分配比JVT平稳且更加接近目标比特。
In this paper,linear R-Q model is applied in the macroblock layer rate control,which can better overcome the problem of parameter oscillation in comparison with quadratic R-Q model.An improved 2D sliding window is proposed for data selection in R-Q model parameter updating.Compared with the JVT-H017 window,this 2D sliding window can excavate the spatial correlation more effectively.A new MAD prediction model is also proposed which has more accurate prediction results than JVT model.Simulation results show that compared with JVT,the peak signal-to-noise ratio increases significantly by using this algorithm,and the bit allocation is relatively stable and closer to the target.
出处
《合肥工业大学学报(自然科学版)》
CAS
CSCD
北大核心
2011年第7期1019-1023,共5页
Journal of Hefei University of Technology:Natural Science
基金
安徽省信息产业厅信息产业发展基金资助项目(2008012)