结合清晰度的无参考图像质量评价  

No-reference Image Quality Assessment Based on Sharpness

作　　者：曹欣 李战明[1] 胡文瑾[2] CAO Xin;LI Zhanming;HU Wenjin(College of Electrical and Information Engineering,Lanzhou University of Technology,Lanzhou 730050;College of Mathematics and Computer Science,Northwest Minzu University,Lanzhou 730000)

机构地区：[1]兰州理工大学电气工程与信息工程学院,兰州730050 [2]西北民族大学数学与计算机科学学院,兰州730000

出　　处：《计算机与数字工程》2020年第4期922-927,共6页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:61561042);西北民族大学“一优三特”学科中央高校基本科研业务费资金项目(编号:31920180117);西北民族大学引进人才基金项目资助

摘　　要：为了提高无参考图像质量评价方法与人眼视觉感知的一致性,论文提出了一种利用待评价图像的清晰度、色度和纹理特征的无参考彩色图像质量评价方法。首先,对待评价图像进行梯度特征以及多尺度能量提取操作得到图像清晰度;然后,进行统一化模式LBP直方图特征提取,另外,还采用了lαβ颜色空间下红绿和蓝黄分量的自然场景统计特征来增强颜色失真的描述,并利用非对称广义高斯模型获得统计参数特征。最后,运用SVR建立图像质量评价模型。利用LIVE数据库中的5种失真类型验证了论文提出的无参考图像质量评价模型的性能。结果表明,论文提出的方法与人眼主观感知具有较好的一致性,并具有复杂度低等优点。In order to improve the consistency of the no reference image quality assessment method and human visual perception,a no reference color image quality assessment method is proposed,which uses the sharpness,chromaticity and texture features of the image to be evaluated.First,the gradient feature of the evaluation image and the multi-scale sub-band energy extraction operation are used to obtain the image clarity.Then,the unified model LBP histogram features are extracted.In addition,the natural scene statistical characteristics of red and green and blue yellow components in the lαβspace are used to enhance the description of color distortion,and the use of Asymmetric generalized Gauss model is used to obtain statistical parameter characteristics.Finally,the model of image quality assessment is established by using SVR.The performance of the no reference image quality assessment model proposed in this paper is verified by 5 distortion types in LIVE database.The results show that the proposed method is consistent with subjective perception of human eyes and has the advantages of low complexity.

关 键 词：多尺度子带能量 清晰度 颜色分量 统一化模式LBP直方图 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]