DVC中基于广义高斯分布的WZ帧重构算法  

Generalized Gaussian Distribution-Based WZ Frame Reconstruction Algoirthm for Distributed Video Coding

作　　者：刘杰平[1] 何越盛 

机构地区：[1]华南理工大学电子与信息学院,广东广州510640

出　　处：《华南理工大学学报（自然科学版）》2014年第8期77-83,共7页Journal of South China University of Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(60972135);北京市"现代信息科学与网络技术"重点实验室和铁道部"铁路信息科学与工程"开放实验室基金资助项目(XDXX1009);广州市科信局对外合作专项资助项目(2012J5100010)

摘　　要：为提高分布式视频编码系统解码WZ(Wyner-Ziv)帧的图像质量,提出了基于广义高斯分布的WZ帧重构算法.该算法充分考虑了边信息与原始WZ帧具有较强相关性的特点,将广义高斯分布作为相关模型;在边信息已知情况下,对给定量化区间、计算WZ帧的条件期望并将其作为WZ重构值;将广义高斯分布的形状参数设为0.5,推导出基于广义高斯分布重构WZ帧的闭式表示.实验结果表明,文中提出的基于广义高斯分布的WZ帧重构算法比最小均方误差(MMSE)重构算法的峰值信噪比提高约0.6 d B,并能有效改善重构WZ帧的图像质量.In order to improve the image quality of the decoded WZ（ Wyner-Ziv） frame in the distributed video coding（ DVC） system,a WZ frame reconstruction algorithm is proposed based on the generalized Gaussian distribution（ GGD）. In this algorithm,the strong correlation between the side information and the original WZ frame is taken into account,and the GGD is used as the correlation model. Then,with the known side information and a given quantization interval,the conditional expectation of the WZ frame is calculated as the reconstructed WZ value. Finally,the closed-form expressions of the reconstructed WZ frames based on the GGD are derived by using a shape parameter of 0. 5. Experimental results indicate that,as compared with the MMSE（ Minimum Mean Square Error） reconstruction algorithm,the proposed algorithm achieves an average PSNR（ Peak Signal to Noise Ratio）increment of about 0. 6 dB and effectively improves the image quality of the decoded WZ frame.

关 键 词：分布式视频编码 WZ重构 广义高斯分布 Wyner-Ziv解码器 

分 类 号：TN919.8[电子电信—通信与信息系统]