摘要
在基本单元层上提出利用峰值信噪比(PSNR)和平均绝对差(MAD)联合估计编码复杂度,并加入了基本单元的位置权值来使目标比特更加精确。由于MAD的线性模型计算复杂度较高,考虑到MAD的时间和空间相关性,本文采用了MAD的加权模型,省略了MAD线性模型的参数更新过程,降低了编码复杂度。试验结果表明:与JM8.6采用的JVT-H017算法相比,本文算法码率精度更高,输出码流更加平稳,图像质量更好。
A new mathod making the target bits more accurate is proposed on basic unit layer,which is used both the peak signal noise ratio and the mean absolute difference to predict the encode complexity besides the location weighted.Because of the high computational complexity of the linear model for MAD prediction and taking into account the space and temporal correlation,a MAD weighted model reducing the computational complexity is presented,which passed over updating the parameters of linear model.The test result shows that,compared with the algorithm using in JVT-H017 by JM8.6,the output has a more accurate bitrate,more stable bit stream and better qulity of pictures.
出处
《电视技术》
北大核心
2009年第S1期45-47,共3页
Video Engineering
