基于HVS特征参数提取的视频质量评价四元数模型  

QUATERNION VIDEO QUALITY ASSESSMENT MODEL BASED ON HVS CHARACTERISTIC PARAMETERS EXTRACTION

作　　者：何叶明 李纲[1] 余凤霞[1] 

机构地区：[1]宁波大学信息科学与工程学院,浙江宁波315211

出　　处：《计算机应用与软件》2014年第7期132-136,140,共6页Computer Applications and Software

基　　金：国家自然科学基金项目(60672072);宁波市自然科学基金项目(2012A610047)

摘　　要：提出一种基于HVS特征参数提取的视频质量评价四元数模型。四元数矩阵从表面纹理信息、帧间残差、亮度边缘和色度边缘四个部分来构建,同时加入HVS感兴趣区域加权和块效应加权。该方法利用四元数矩阵奇异值分解提取特征向量的方式来获取视频的特征信息。通过VQEG Phase I测试数据集的实验结果证明,该方法在非线性回归后相关系数、斯皮尔曼相关系数、离出率等指标均优于传统的其他视频质量评价算法,有效地提高了视频质量评价的主客观一致性。We propose a quaternion video quality assessment model which is based on HVS characteristic parameters extraction. The quaterni0n matrix is constructed with four components including the surface texture information, interframe residuals, luminance edge and chrominance edge. Meanwhile, the HVS interested area weight and blocking-effect weight are also affiliated. The method makes use of approach which extracts the eigenvector by using quaternion matrix singular value decomposition to acquire video＇ s feature information. It is proved by the experimental results on the VQEG Phase ! test dataset that this method outperforms other video quality assessment algorithms in indices of the correlation coefficient after the nonlinear regression, Spearman correlation coefficient and out-rate, etc. , and effectively improves the objective and subjective consistency in video quality assessment.

关 键 词：视频质量评价 四元数矩阵 奇异值分解 块效应 视频质量专家组 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]