Wyner-Ziv视频编码中无反馈速率控制算法研究  

No-feedback rate control algorithm for Wyner-Ziv video coding

作　　者：宋彬[1] 杨明明[1] 秦浩[1] 贺红[1] 

机构地区：[1]西安电子科技大学ISN国家重点实验室,陕西西安710071

出　　处：《通信学报》2011年第12期1-7,共7页Journal on Communications

基　　金：国家自然科学基金资助项目(60802032);高等学校创新引智基地基金资助项目(B08038);长江学者和创新团队发展计划基金资助项目(IRT0852)~~

摘　　要：为了避免在分布式视频编码系统中使用反馈信道,提出了一种基于Wyner-Ziv编码的无反馈速率控制算法。首先,利用目标码率和目标帧率进行GOP层码率分配;然后,根据原始图像的帧间相关性动态选择量化因子和量化矩阵来分配每个GOP内关键帧和Wyner-Ziv帧的比特数;接下来,利用系数带级的相关性计算相关噪声模型参数,并选择对应的LDPC校验矩阵,提出Wyner-Ziv帧的无反馈比特面速率控制算法。实验结果表明,在给定目标码率下,所提算法的编码码率误差小于0.57%,且与现有无反馈速率控制算法相比,解码恢复图像的PSNR(峰值信噪比)可以提高1dB。另外,该算法基本没有增加编码端复杂度,可用于实际分布式视频通信系统。To avoid using feedback channel, an encoder rate control algorithm for Wyner-Ziv video coding was proposed. Firstly, using the target frame-rate and the target bit-rate, It could determine the bit allocation for each group-of-picture （GOP）. According to the correlation between adjacent frames, the bit budget for a GOP were further allocated to key frame and Wyner-Ziv frames within the GOP by dynamically choosing quantization parameter and quantization matrices. Then, with refinement of correlation-noise model （CNM） parameter, different LDPC （low-density parity-check） check matrices might be chosen for each bitplane to control the encoding bitrate of Wyner-Ziv frame. Experimental results show that the difference between expected target bitrate and actual encoded bitrate is less than 0.57% using the proposed method, and compared with existing algorithm, average PSNR （peak signal-to-noise ratio） of decoded video quality can be increased by 1 dB. In addition, our algorithm slightly increases encoder complexity, and can be used in practice.

关 键 词：WYNER-ZIV编码 无反馈 速率控制 相关噪声模型 

分 类 号：TN919.8[电子电信—通信与信息系统]