融合显著图和保真图的全参考图像质量评价  

Full-reference image quality assessment based on saliency and fidelity maps

作　　者：郭迎春[1] 于洋[1] 师硕[1] 于明[1] 

机构地区：[1]河北工业大学计算机科学与软件学院,天津300401

出　　处：《光电子．激光》2016年第11期1228-1237,共10页Journal of Optoelectronics·Laser

基　　金：天津市科技计划项目(14RCGFGX00846);河北省自然科学基金(F2015202239)资助项目

摘　　要：提出一种融合显著图(SM)和保真图(FM)的全参考图像质量评价算法,用于评价质降图像的失真度。利用亮度和色度的相似度提取质降图像相对于参考图像的FM;对参考图像进行区域划分、全局显著性提取和纹理边缘补充得到SM,将SM与质降图像的FM融合得到基于感知的显著保真图(PSM),计算质降图像的客观评价得分。在标准数据库上的实验结果表明,本文方法与主观评价能够很好保持一致,并对LIVE图像库中的5种失真图像均有很好的表现。Considering the fact that human eyes pay more attention to the salient regions of images,this paper proposes a full-reference image quality assessment based on salient region map a nd fidelity map.Firstly,the fidelity map of distorted image is obtained by calculating the similarity betwee n the distorted and undistorted images on brightness and chromaticity components;then,the salient map of the u ndistorted image is obtained by segmenting regional blocks,extracting the global saliency of each block and complementing the edge texture of local feature;finally,using the saliency map as weight,the objective asses sment score is obtained by calculating the weighted fidelity map.The experimental results in LIVE image dat abase show that the image quality assessment method based on saliency detection proposed here has good performanc e on assessing the five kinds of distorted images.

关 键 词：全参考图像质量评价 显著图(SM) 感知保真图(PSM) 客观评价 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]