Abstract: Due to limited bandwidth of communication channel，it is necessary to control the bit rate of video coding output.This paper proposes the estimation of initial value of QP and macroblock layer rate control.In the computation of initial value of QP，this paper proposes bpp，entropy and gradient as an estimate of the complexity of the image.It calculates the initial value of QP through this way.Meanwhile，this paper improves the prediction of MAD by using motion vector and the MAD gradient.Thus，it can achieve the control of the bit rate more effectively.The simulation results show that compared with the algorithms of H.264 reference software JM11.0，the proposed algorithm enhances the PSNR and is closer to the target bit.

Key words: H.264, rate control, Quantization Parameter（QP）, Mean Absolute Difference（MAD）

摘要： 由于通信信道带宽有限，有必要对视频编码的输出码率进行控制。基于视频编码标准H.264采用的码率控制算法，在初始QP值的设定以及宏块层码率控制方面提出了改进。在初始QP值的计算上，提出了利用基于bpp、熵和梯度的图像复杂度对初始QP值进行估计的方法。另外利用时间上的预测信息和空间上的MAD梯度信息在宏块层对MAD预测进行了改进，能更有效地实现对码率的精确控制。实验结果表明，与H.264标准参考软件JM11.0中的算法相比，改进后的算法比测试序列在峰值信噪比（PSNR）上得到了提高，同时能更接近目标比特。

关键词: H.264, 码率控制, 量化参数, 平均绝对值差