光学精密工程2025，Vol.33Issue(18)：2944-2961,18.DOI:10.37188/OPE.20253318.2944
基于图像画面信息及其视觉感知的无参考图像质量评价
No reference image quality assessment based on image tableau information and their visual perception
摘要
Abstract
It aimed to propose an image quality assessment(IQA)method that conformed to visual per-ception and had high comprehensive benefits of accuracy,generalization,and complexity,which would better control image processing and meet its practical application needs.Based on the image tableau infor-mation and their features such as brightness,chromaticity,texture,clarity,and local contrast,and consid-ering human perception effects such as contrast sensitivity,non-linear perception of brightness,visual com-fort,and texture complexity perception,a no-reference IQA method,namely BCTCSP,was proposed.In BCTCSP,firstly,by analyzing the relationship between image quality and image brightness,grayscale distribution,color depth and saturation,non-linear perception of brightness,and visual perception com-fort,a quantitative and computational method was proposed to obtain the contribution and impact of image brightness,chromaticity,and visual perception on IQA.Then,combining the gray-gradient co-occur-rence matrix to calculate and statistically analyze image texture features,and using texture weighted aver-aging and the HVS complex object perception model,a method was proposed to quantify and calculate the contribution and impact of image texture information entropy and its visual perception on IQA.Next,the contrast value and detection threshold of each point in the image were calculated,subsequently combining the contrast sensitivity characteristics and their models,and masking properties of HVS,the contribution and impact of local contrast and visual perception of the image on IQA were quantified and calculated.Af-terwards,four factors including sharpness,signal-to-noise ratio,proportion of high-frequency compo-nents,and resolution,were used to describe the clarity of the image,and their quantification and calcula-tion methods were proposed to obtain the clarity index of the image.Finally,synthesizing four factors,an IQA model was constructed,and its measurement standards were quantified.Meanwhile,6 430 distorted images from 6 open databases(TID2013,CSIQ,LIVE,IVC,SPAQ,and Koniq-10k)were tested and verified,and in terms of accuracy,complexity,generalization,and their comprehensive benefits,BCTC-SP was compared with 28 existing and typical IQA models.The experimental results show that the accura-cy PLCC of the proposed model reaches a minimum of 0.892 1,a maximum of 0.966 4 among the 6 data-bases,and the weighted PLCC of 6 databases reaches 0.917 4.Its comprehensive benefits are higher than those of the 28 existing IQA models.The comprehensive results indicate that the proposed model is an ef-fective and high-performance NR-IQA model.关键词
图像质量评价/图像画面/视觉感知/纹理信息熵Key words
image quality assessment/image tableau/visual perception/texture information entropy分类
计算机与自动化引用本文复制引用
姚军财,申静..基于图像画面信息及其视觉感知的无参考图像质量评价[J].光学精密工程,2025,33(18):2944-2961,18.基金项目
国家自然科学基金(No.61301237) （No.61301237）
江苏省自然科学基金面上项目(No.BK20201468) （No.BK20201468）
江苏省高校"青蓝工程"中青年学术带头人资助项目(苏教师函[2022]29号)和南京工程学院创新基金重大项目(No.CK-JA202304) （苏教师函[2022]29号）
