双尺度画面细节信息引导的全参考图像质量客观评价OA

In industrial vision systems,full-reference image quality assessment(FR-IQA)has emerged as a major research focus.To address this problem,a novel FR-IQA model,termed a dual-scale tableau de-tail-guided model,is proposed.First,the test image is transformed into a specific color space to decouple luminance and chrominance channels.Subsequently,a luminance gradient similarity component is con-structed based on gradient interactions among the reference image,the distorted image,and their fused representation.This component is further integrated with a chrominance orientation consistency analysis to derive a color similarity measure.Meanwhile,a dual-scale tableau detail accumulation component is devel-oped by fusing spectral residual,edge,and adaptive contrast features of the luminance channel,enabling effective characterization of cumulative detail information.Finally,the standard deviation features of these components are weighted and aggregated through feature coefficients to produce the overall quality score.The reliability of the proposed model is validated on the LIVE,CSIQ,TID2008,TID2013,and KA-DID-10K databases using four evaluation metrics:Spearman rank-order correlation coefficient(SROCC),Pearson linear correlation coefficient(PLCC),Kendall rank-order correlation coefficient(KROCC),and root mean square error(RMSE).Experimental results indicate that the PLCC ranges from 0.876 8 on KADID-10K to 0.967 8 on LIVE,while the SROCC varies from 0.864 8 on TID2013 to 0.961 0 on CSIQ.The proposed model demonstrates superior computational efficiency compared with state-of-the-art and deep learning-based FR-IQA methods,while maintaining robust and generalizable predictive perfor-mance.