摘要
在H.264码率控制算法中,预测基本单元平均绝对差值(MAD)时只考虑时间相关性,分配基本单元比特数时仅考虑MAD因素。针对该问题,提出采用时间线性预测模型与时空加权预测模型相结合的方法,进行基本单元MAD预测,并根据前一基本单元的MAD预测误差,对当前基本单元MAD预测值进行修正。同时,采用峰值信噪比(PSNR)与MAD相结合的方法,估计基本单元复杂度,进而分配基本单元目标比特数。仿真结果表明,与JM8.6模型相比,改进算法在提高码率控制精度的同时,PSNR平均可提高0.1 dB。
In view of the limitation caused by considering time correlation only in Mean Absolute Difference(MAD) prediction and MAD factor only in bit allocation of basic unit in Rate Control(RC) algorithm for H.264, this paper proposes an improved algorithm in which the predicted MAD is calculated by a combination of linear prediction model and the space-time weighted prediction model, and corrected according to the prediction error of the previous basic unit. Besides, the basic unit complexity is estimated by a combined Peak Signal Noise Ratio(PSNR)-based and MAD-based method, based on which bit allocation is done. Simulation results show that compared with JMS.6, the proposed algorithm not only can make the output bit rate more accurate, but also can improve the average PSNR by 0.1 dB.
出处
《计算机工程》
CAS
CSCD
2013年第10期275-278,共4页
Computer Engineering
基金
国家"863"计划基金资助项目(2011AA01A105)
上海市科委基金资助项目(10DZ1500700
11DZ1500500)
关键词
H
264标准
码率控制
基本单元
平均绝对差值
峰值信噪比
基本单元复杂度
H.264 standard
Rate Control(RC)
basic unit
Mean Absolute Difference(MAD)
Peak Signal Noise Ratio(PSNR)
basic-unit complexity