高效率视频编码中基于R-λ模型的码率控制优化算法

Improved rate control algorithm based on R-λ model in high efficiency video coding

针对高效率视频编码(High Efficiency Video Coding,HEVC)在编码过程中复杂度较高的最大编码单元(Largest Coding Unit,LCU)层带宽相对缺乏而复杂度较低的LCU层带宽相对充裕的问题,为了改善编码性能,本文提出一种基于R-λ模型的码率控制改进算法。首先,采用梯度(Gradient,G)与平均绝对差(Mean Absolute Difference,MAD)加权表示LCU层图像复杂度。然后,利用已编码单元的反馈信息进一步调整量化参数。最后,利用视频的编码失真结合牛顿法对模型参数进行更新。实验结果表明,与现有的HEVC码率控制算法相比,本文算法的编码性能更加优良并且具有较强的鲁棒性,即:与HEVC标准的码率控制算法相比,改进算法的平均码率误差下降0.055%,平均亮度分量峰值信噪比提高了0.07dB;与参考文献的算法相比,改进算法的平均码率误差下降0.023%,平均亮度分量峰值信噪比提高了0.05dB。

In order to improve coding performance of high efficiency video coding （HEVC）, this paper proposes an improved rate control algorithm based on R-λ model. Firstly, it utilizes gradient and mean absolute difference （MAD） weighting to represent the complexity of the gradient LCU layer image. Then,it employs the feedback information of the encoded units to further adjust the quantization parame- ters. Finally, the model parameters are updated by utilizing video coding distortion combined with Newton method. The experimental results show that the proposed method is better in coding performance and ro- bustness than the current HEVCPs rate control algorithms. The average bit-error of the improved algo- rithm is reduced by 0.055%, the average luminance peak to noise ratio is increased by 0.07 dB compared with rate control algorithm for the HEVC standard, the average bit-error d the improved algorithm is reduced by 0.023 %, and the average luminance peak to noise ratio is increased by 0. 05 dB compared with the reference algorithms.