北京邮电大学学报

  • EI核心期刊

北京邮电大学学报 ›› 2015, Vol. 38 ›› Issue (1): 103-107.doi: 10.13190/j.jbupt.2015.01.020

• 研究报告 • 上一篇    下一篇

像素域基于广义高斯分布的WZ帧重构方案设计

刘杰平, 何越盛, 韦岗   

  1. 华南理工大学 电子与信息学院, 广州 510640
  • 收稿日期:2014-04-10 出版日期:2015-02-28 发布日期:2015-03-30
  • 作者简介:刘杰平(1961—),女,副教授,硕士生导师,E-mail:eeliujp@scut.edu.cn.
  • 基金资助:

    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项目(2012J5100010)

Design of WZ Frame Reconstruction Technology Based on Generalized Gaussian Distribution in Pixel-Domain

LIU Jie-ping, HE Yue-sheng, WEI Gang   

  1. School of Electronic and Information Engineering, South China University of Technology, Guangzhou 510640, China
  • Received:2014-04-10 Online:2015-02-28 Published:2015-03-30

摘要:

为提高分布式视频编码系统解码WZ帧的图像质量,提出了像素域基于广义高斯分布的WZ帧重构方案. 该方案充分考虑了边信息与原始WZ帧间的相关性,以广义高斯分布作为虚拟相关信道模型,提高率失真性能;用广义高斯分布做相关噪声模型,对给定量化区间,计算边信息已知情况下WZ的条件期望作为WZ重构值;为了不过多地增加重构方案的复杂度,将广义高斯分布的形状参数固定为0.5,推导出重构WZ帧的闭式表示. 实验结果表明,基于广义高斯分布的WZ帧重构方案能有效提高率失真性能和改善重构WZ帧的图像质量.

关键词: 分布式视频编码, 相关噪声模型, 广义高斯分布, 重构

Abstract:

In order to improve the image quality of the decoded WZ (Wyner-Ziv) frame in distributed video coding system,WZ frame reconstruction technology based on generalized Gaussian distribution (GGD) was proposed. The correlation between the side information and the original WZ frame was considered as well. GGD was used for virtual correlation channel model so as to improve the rate distortion (RD) performance. Meanwhile,the condition expectation of the WZ was computed as the reconstructed WZ value for the given quantization interval and the known side information in which GGD is used as correlation noise model. The shape parameter of GGD is fixed to 0.5 so that it won't add extra complexity to the reconstruction technology. The closed-form expression of the optimal reconstructed values was derived. Experiments indicate that the proposed WZ frame reconstruction technology can improve the RD performance and image quality of the decoded WZ frame.

Key words: distributed video coding, correlation noise model, generalized Gaussian distribution, reconstruction

中图分类号: