基于平均预测残差的H.264宏块层码率控制算法

针对H.264码率控制模型的“蛋鸡悖论”，提出一种用平均预测残差来表示当前宏块编码复杂度的算法。首先根据各个宏块不同的编码复杂度进行准确的比特分配,然后根据改进的二次R-D预测模型来计算宏块的量化参数，并进行率失真优化,最后进行码率控制模型系数的更新。将算法在JM8.6平台上进行实验，结果显示，在相同码率下，本文算法的平均控制精度比JVT-H017提高0165 kb/s，平均PSNR比JVT-H017提高013 dB，表明本算法能够更准确控制码率并获得更高质量的主观视觉。

H.264 macro-block layer rate control algorithm based on average prediction residual

In order to solve the problem of “Egg-Chichen Absurdity” in the rate control model H.264, this paper presents an algorithm which can express the coding complexity of current macro-block using the average prediction residual. Firstly, a more precise bit allocation scheme is implemented according to the coding complexity of each macro-block, and then the quantization parameters of macro-block are calculated using the improved quadratic Rate-Distortion(R-D) prediction model and R-D optimization is achieved. Finally, the coefficients of the rate control model have been updated. This algorithm is implemented in JM8.6, and the experimental results show that the proposed method can improve 0.165 kb/s than JVT-H017 in the control accuracy, and improve 0.13 dB in PSNR on the average at the same rate. It can be concluded that the proposed method can control the rate more accurately and can obtain a higher quality subjective vision.