基于人眼视觉系统的H.264/AVC码率控制算法  

H.264/AVC rate control algorithm based on human visual system

作　　者：孙乐[1,2] 戴明[1] 陈晓露[1,2] 王子辰[1,2] 赵春蕾[1,2] 

机构地区：[1]中国科学院长春光学精密机械与物理研究所航空光学成像与测量重点实验室,吉林长春130033 [2]中国科学院大学,北京100039

出　　处：《光学精密工程》2014年第4期1020-1028,共9页Optics and Precision Engineering

基　　金：国家863高技术研究发展计划资助项目(No.2008AA121803);国家973重点基础研究发展规划资助项目(No.2009CB72400603B)

摘　　要：提出了一种基于人眼视觉系统(HVS)特性的H.264/AVC码率控制方案以解决传统的H.264/AVC码率控制算法中由于码率分配不当造成视频质量较差的问题。该方案利用人眼视觉系统对视频场景感知的选择性设计了一种新的基于区域的码率控制算法。首先,利用当前帧和前一帧的信息快速地检测出运动区域和非运动区域。其次,根据宏块所在区域的复杂度进一步将非运动区域细分为平坦区域和复杂区域。最后,对这3种区域依据不同的视觉敏感度权重分配适当的比特数。基于不同区域峰值信噪比(PSNR)、连续帧PSNR、主观评价以及缓冲器充盈度综合评价了本文算法的性能,结果表明,提出的基于区域的码率控制方案对不同视觉特性区域的比特分配有较高的鲁棒性,与JVT-G012的码率控制算法相比能获得较好的主观和客观编码质量。This paper proposes a rate-control scheme based on Human Visual System（HVS） for H.264/AVC to improve the poor video quality caused by the normal rate control in traditional H.264/AV scheme.A novel rate control algorithm was designed based on selectivity to visual scene in the perception of the HVS.Firstly,the moving region and non-moving region were quickly detected by using the information from a current frame and a previous frame.Then,non-moving region was sub-di-vided into complex regions and flat regions according to the complexity of MBs region.Finally,the appropriate bits were allocated onto these three regions based on their own perceptual sensitivity weighting factors and the performance of the algorithm was evaluated by Peak Signal to Noise Ratios （PSNRs） in different regions and continuous frames,subjective evaluation and filling degree of a bumper.Experimental results show that the proposed region-based rate control scheme has high ro bustness in bit allocation in different regions.In comparing with the JVT-G012 in H.264 rate control approach,it can achieve better subjective and objective quality.

关 键 词：H 264 AVC 码率控制 人眼视觉系统 运动区域 码率分配 

分 类 号：TN941.1[电子电信—信号与信息处理] TP391[电子电信—信息与通信工程]