一种运动矢量和DCT融合的无参考视频质量评价模型  

An Associated Quality Assessment Model of Motion Vector and DCT without Reference Video

作　　者：吴丽丽[1] 余春艳[1] 陈国龙[1] 郑维宏 

机构地区：[1]福州大学数学与计算机科学学院,福建福州350108 [2]福建星网视易信息系统有限公司,福建福州350002

出　　处：《图学学报》2014年第3期442-449,共8页Journal of Graphics

基　　金：国家自然科学基金资助项目(60805042);福建省自然科学基金资助项目(2010J01329;2011J05150;2012J01262;2013J01231);福建省重大产学合作基金资助项目(2010H6012;2011H6014)

摘　　要：网络丢包引发的视频失真是网络视频流媒体服务质量评价的研究重点。网络丢包引发的人眼视觉感受与丢包所在帧类型、视频帧的运动剧烈状态和纹理丰富程度有着显著关联,提出了一种运动矢量和DCT融合的无参考视频质量评价模型,以发生丢包的宏块为基本研究对象,以运动矢量表征其运动剧烈程度,以DCT直流系数刻画其纹理丰富程度,无需获取原始视频任何信息,无需完全解码,关联运动矢量和DCT完成视频质量的客观评价。实验结果表明,该模型的评价结果与主观评价结果有较好的一致性。Network video distortion caused by packet loss is a research priority of online video streaming service quality evaluation. An associated quality assessment model of motion vector and DCT without reference video is provided. This model associates the human visual perception caused by network packet loss with the frame type whose packet is lost, the movement state and the richness of texture. This model takes the macro blocks that are lost as the basic study object. And it calculates the motion intenseness by using the macro blocks? motion vector and calculates the richness of its texture by using the DCT coefficients. And this model evaluates videos?quality objectively by associating the motion vector and DCT without any information of the original video and doesn?t need decoding completely. Experimental results show that this model's evaluation results have a good consistency with the subjective evaluation results.

关 键 词：无参考视频质量评价 丢包 运动剧烈程度 DCT直流系数 空域复杂度 

分 类 号：TP918.8[自动化与计算机技术]