基于局部信息失真建模的图像质量评价方法  

Image quality assessment based on local information distortion modeling

作　　者：卢彦飞[1,2,3] 张涛[1] 郑健[3] 李铭[1,2,3] 章程[1,2,3] 

机构地区：[1]中国科学院长春光学精密机械与物理研究所,吉林长春130033 [2]中国科学院大学,北京100049 [3]中国科学院苏州生物医学工程技术研究所,江苏苏州215163

出　　处：《激光与红外》2015年第8期987-993,共7页Laser & Infrared

基　　金：国家自然科学基金项目(No.61201117;No.61301042);国家重大科学仪器设备开发专项(No.2011YQ040082);国家科技支撑计划(No.2012BA113B04);江苏省自然基金项目(No.BK2012189);苏州市科技计划项目(No.ZXY2013001)资助

摘　　要：针对传统的基于像素差值统计的方法以及结构相似度方法不能很好地反映主观评价结果的情况,提出了一种利用图像局部信息失真建模的质量评价方法。该方法通过考虑人眼视觉系统的特点,对像素灰度失真、局部对比度失真和局部结构失真进行建模,并利用局部方差作为权重,得到了最终的图像质量评价测度。其物理意义明确,而且计算相对简单。在LIVE图像数据库上的实验表明,本文方法对于jp2k,jpeg,gblur和fastfading失真的质量预测准确性和一致性都很高,均优于结构相似度方法,对于wn失真也有较好的预测结果。与几种公认较好的方法相比,本文方法表现出了很好的预测性能,得到了与人眼主观感知更加一致的结果。As traditional pixel -difference statistics and structural similarity （SSIM）cannot well reflect subjective e-valuation results,a novel image quality assessment method based on local information distortion modeling is pro-posed.According to some characteristics of human visual system （HVS）,models are established for the pixel gray dis-tortion,local contrast distortion and local structure distortion.Taking local variance as weight,the final image quality assessment metric is obtained.The experiments on the LIVE database indicate that the proposed method has high pre-diction accuracy and consistency on distortion types like jp2k,jpeg,gblur and fastfading,and it is better than the SSIM method.Also,a fairly good prediction result is achieved for wn distortion type.Compared with other assessment meth-ods,the proposed method has good predictive performance,and its calculation is quite simple,and it is more consistent with subjective evaluation.

关 键 词：图像质量 结构相似度 局部二值模式 局部方差 局部失真 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]