计算机工程

• 开发研究与工程应用 • 上一篇    下一篇

稀疏统计二次Renyi熵的运动估计匹配准则

邓志超,汪同庆   

  1. (重庆大学光电技术及系统教育部重点实验室,重庆400044)
  • 收稿日期:2013-07-01 出版日期:2014-08-15 发布日期:2014-08-15
  • 作者简介:邓志超(1986-),男,硕士研究生,主研方向:视频编码技术;汪同庆,教授、博士生导师。
  • 基金项目:
    国家科技支撑计划基金资助项目(2007BAG06B06);中央高校基本科研业务费专项基金资助项目(10611201312014)。

Matching Criterion Using Sparse Statistics on Quadratic Renyi’s Entropy for Motion Estimation

DENG Zhi-chao,WANG Tong-qing   

  1. (Key Laboratory of Optoelectronic Technology & System,Ministry of Education,Chongqing University,Chongqing 400044,China)
  • Received:2013-07-01 Online:2014-08-15 Published:2014-08-15

摘要: 运动估计是视频编码技术的核心,通过对运动估计基本原理的分析可知匹配准则是运动估计的关键环节。针对传统匹配准则描述块匹配精度不高,导致信息冗余较多的不足,提出一种稀疏统计二次Renyi熵的运动估计匹配准则。该匹配准则在计算Renyi熵时,引入了统计直方图,采用统计的方式计算概率密度函数,并结合基于梯度的图像质量评价和运动矢量中心偏离特性,对直方图的统计区间进行稀疏化。实验结果表明,该匹配准则简化了Renyi熵概率密度函数的计算,通过统计区间的稀疏化减少了80%以上的乘法运算量,对运动剧烈的视频序列能够得到优于传统绝对误差和函数的峰值信噪比,取得更好的图像质量。

关键词: 运动估计, 匹配准则, 二次Renyi熵, 稀疏统计, 统计直方图, 概率密度函数

Abstract: Motion estimation is the key of video coding,with the analysis of the principle of motion estimation,it can be obtained that matching criterion is the key link.In view of the traditional matching criteria,the block matching accuracy is not high,which leads to more insufficient information redundancy,it proposes a new matching criterion using sparse processing on Quadratic Renyi’s Entropy(QRE) for motion estimation.When calculating the Renyi entropy,the matching criterion introduces statistical histogram to calculate the probability density function,and according to the theory of image quality assessment based on gradient motion vector and the center deviation characteristics,it makes the statistical interval of its histogram sparse.Experimental results show that,this matching criterion simplifies the calculation of probability density function,multiplication computation is greatly reduced by more than 80%with sparse histogram statistics,and in view of the dramatic video sequence,the matching criterion movement can be superior to traditional sum of absolute difference function in peak signaltonoise ratio and image quality.

Key words: motion estimation, matching criterion, Quadratic Renyi’s Entropy(QRE), sparse statistics, statistical histogram, probability density function

中图分类号: