摘要
基于高效视频编码标准的x265编码器根据图像复杂度来分配比特,复杂图像往往包含运动变化较大的高频信息,其时域相关性较弱且消耗较多比特,导致分配给运动变化平缓图像的比特减少,进而影响编码质量且码率波动较大。同时,x265编码器采用独立率失真优化技术编码,忽略了编码单元间在时域上的相关性,进而损失编码性能。针对上述问题,提出一种基于时域依赖的编码树单元级码率控制算法。首先,根据迭代策略寻找最合适的量化参数进行帧级比特分配;其次,建立零延时的失真时域反向传播模型并计算失真影响因子;最后,将失真影响因子用于调整编码单元的拉格朗日乘子及量化参数。实验结果显示,相较于x265-3.6的码率控制算法,所提算法的平均BD-rate码率节省达到5.6%。
The x265 encoder,based on the efficient video coding standard,allocates bits according to the complexity of the image.Complex images often include high-frequency information with significant motion changes,characterized by weak temporal correlation and high bit consumption.This results in a reduction of bits assigned to images with smooth motion changes,consequently affecting encoding quality and causing significant bitrate fluctuations.Additionally,the x265 encoder employs independent rate-distortion optimization techniques for encoding,neglecting the temporal correlation between encoding units and consequently leading to a loss in encoding performance.To address these issues,this paper proposed a temporal-dependent coding tree unit-level rate control algorithm.Firstly,it explored the most suitable quantization parameters for frame-level bit allocation based on an iterative strategy.Secondly,it established a distortion temporal back-propagation model and calculated distortion impact factors.Lastly,it used these distortion impact factors to adjust the Lagrange multipliers and quantization parameters of coding units.Experimental results demonstrate that compared to the rate control algorithm of x265-3.6,the proposed algorithm achieves an average BD-rate savings of 5.6%.
作者
程宝平
陶晓明
黄敏峰
谢小燕
杜金
杨栩
Cheng Baoping;Tao Xiaoming;Huang Minfeng;Xie Xiaoyan;Du Jin;Yang Xu(Dept.of Electronic Engineering,Tsinghua University,Beijing 100084,China;China Mobile(Hangzhou)Information Technology Co.,Ltd.,Hangzhou 311100,China;School of Information&Communication Engineering,University of Electronic Science&Technology of China,Chengdu 611731,China;School of Physics&Engineering Technology,Chengdu Normal University,Chengdu 611130,China)
出处
《计算机应用研究》
CSCD
北大核心
2024年第5期1489-1495,共7页
Application Research of Computers
基金
国家自然科学基金资助项目(62020106011,U19A2052,62171257,U22B2001)。