采用内容划分方法的视频质量评价  

Video quality assessment using content-partitioned approach

作　　者：姚杰[1,2] 谢永强 谭建明[2] 李东[1,3] 唐超[2] 王伏华[2] 

机构地区：[1]中国电子设备系统工程公司,北京100141 [2]重庆通信学院军事网格实验室,重庆400035 [3]解放军理工大学指挥自动化学院,南京210007

出　　处：《计算机工程与应用》2013年第11期158-161,231,共5页Computer Engineering and Applications

基　　金：国家创新基金(No.11c26215115768);重庆市重点攻关基金项目(No.cstc2011ab2064)

摘　　要：目前基于结构相似性的图像质量评价算法均是对图像进行整体质量分析,但图像中不同的区域存在着不同的结构特性和视觉感知特性,而对图像进行整体质量分析无法有效反应出这些差异。鉴于此,提出了一种基于内容划分的结构相似性图像质量评价算法,根据图像不同区域的变化率将图像分为4个部分,分别进行质量评价。采用运动估计的帧加权的方式将该方法扩展到视频质量评价中。实验证明了该算法与目前比较流行的几个算法相比具有较高的评价准确性。Current structural similarity based image quality assessment algorithm is generally the overall image quahty analysis. However, different regions in image have different structural characteristics and visual perceptions, and the overall quality analy- sis can not reflect these differences effectively. In this view, a content-partitioned structural similarity image quality assessment algorithm is presented, which partitions an image into four regions according to their different gradient magnitudes and assesses the qualities of these regions respectively. A frame motion estimation weighted approach is used to extend this approach to video quality assessment. The experiments show that the proposed is more accurate than several modem popular algorithms.

关 键 词：图像质量评价 视频质量评价 结构相似性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]