华南理工大学学报(自然科学版) ›› 2012, Vol. 40 ›› Issue (4): 64-70,78.

• 计算机科学与技术 • 上一篇    下一篇

双域无参考视频质量评价算法

林翔宇 陈耀武   

  1. 浙江大学 数字技术及仪器研究所,浙江 杭州 310027
  • 收稿日期:2011-11-08 修回日期:2012-01-31 出版日期:2012-04-25 发布日期:2012-03-01
  • 通信作者: 林翔宇(1983-) ,男,博士生,主要从事视频编解码及质量评价研究. E-mail:linxiangyu@zju.edu.cn
  • 作者简介:林翔宇(1983-) ,男,博士生,主要从事视频编解码及质量评价研究.
  • 基金资助:

    国家科技支撑计划项目( 2009BAF39B03) ; 国家自然科学基金仪器专项资助项目( 40927001) ; 浙江大学中央高校基本科研业务费专项资金资助项目

A Two-Domain No-Reference Video Quality Assessment Algorithm

Lin Xiang-yu  Chen Yao-wu   

  1. Institute of Advanced Digital Technology and Instrumentation,Zhejiang University,Hangzhou 310027,Zhejiang,China
  • Received:2011-11-08 Revised:2012-01-31 Online:2012-04-25 Published:2012-03-01
  • Contact: 林翔宇(1983-) ,男,博士生,主要从事视频编解码及质量评价研究. E-mail:linxiangyu@zju.edu.cn
  • About author:林翔宇(1983-) ,男,博士生,主要从事视频编解码及质量评价研究.
  • Supported by:

    国家科技支撑计划项目( 2009BAF39B03) ; 国家自然科学基金仪器专项资助项目( 40927001) ; 浙江大学中央高校基本科研业务费专项资金资助项目

摘要: 为了提高视频质量评价的精确度,降低计算复杂度,提出了一种双域无参考视频质量评价算法. 在压缩域,利用从码流中提取的编码信息,结合视频内容复杂度来计算视频相似度; 在像素域,结合人类视觉系统的特性,通过块效应和模糊效应评估来计算视频失真度; 最后,综合视频相似度和失真度得出视频客观质量. 文中选用LIVE Video Quality Database 对算法效果进行了评估,结果表明: 该算法的预测精确性和单调性分别达到0.8013 和0.7861,优于全部10 种对比的视频客观质量评价方法; 同时,该算法具有较低的计算复杂度,可以满足实时应用要求.

关键词: 视频质量, 无参考评价算法, 视频相似度, 视频失真度, 人类视觉系统

Abstract:

In order to improve the accuracy and reduce the computational complexity of video quality assessment,a two-domain no-reference video quality assessment algorithm is proposed. In this algorithm,first,encoding information is extracted from the bitstream and video similarity is calculated according to the complexity of video contents in the compressed domain. Then,by taking into consideration the characteristics of human visual system,the video distortion is calculated by measuring the influence of blockiness and the blur artifacts in the pixel domain. Thus,the video quality is assessed according to the video similarity and the video distortion. The proposed algorithm is finally tested on the LIVE Video Quality Database. The results show that the proposed algorithm,which possess a prediction accuracy and a monotonicity respectively up to 0.8013 and 0.7861,is better than the existing ten video quality assessment techniques,and that it meets the requirements of real-time application because it is of low computational complexity.

Key words: video quality, no-reference assessment algorithm, video similarity, video distortion, human visual system