结合视觉感知和结构相似性的图像质量评价

Image Quality Assessment Combined with Visual Perception and Structural Similarity

目的研究一种精度、泛化性和复杂性综合效益较高的图像质量评价(IQA)模型。方法结合图像失真特征和人眼空间感知特性,基于结构相似性模型SSIM,提出一种基于图像结构感知相似性的IQA方法及其模型ICBS。该方法首先结合人眼对比敏感、频率敏感、亮度感知和掩蔽等特性及其模型,提出一种人眼对真实图像感知的数学模型;然后,模拟人眼对真实图像的感知,通过该感知模型对图像进行处理,以降低图像的视觉冗余,获得感知图像;最后,基于SSIM的建模方式,结合感知图像的特征,提出ICBS模型。结果为了说明所提模型的性能,采用3个开源数据库中35种失真类型的4645幅失真图像,从整体评价和单独评价每一失真类型2个方面进行了仿真测试,结果表明,ICBS的精度在3个数据库中,其加权PLCC值达到0.917,SROCC值达到0.912。基于精度、泛化性能和复杂性的综合效益,与7个现有典型模型的评价结果进行了对比分析,结果表明,ICBS的精度比SSIM的精度在3个数据库中平均提高了10.77%,且其性能综合效益优于7个现有模型。结论综合实验结果和理论分析结果可知,所提模型是一种有效、性能优异的IQA模型。

The work aims to develop an image quality assessment(IQA)model with high comprehensive benefits in terms of accuracy,generalization and complexity.Combined with image distortion characteristics and human spatial visual system(HVS)characteristics and based on the structural similarity index(SSIM)model,an IQA method and model based on the image structure perception similarity(ICBS)was proposed.In the method,firstly,a mathematical model for human detecting the real images was proposed combined with the characteristics of the contrast sensitivity,frequency sensitivity,luminance nonlinearity and masking of HVS.Then,the perception of human eyes to the real image was simulated and the real images were processed with the proposed image perception model to reduce visual redundancy.Thus,the perceptual images were obtained.Finally,based on the modeling method of SSIM and combined with the characteristics of perceptual image,an IQA model(namely ICBS)was proposed.Further,in order to illustrate the performance of ICBS,4645 distorted images of 35 distortion types in three databases were used for simulation from two aspects of overall evaluation and IQA for each distortion type.The results showed that the accuracy of ICBS was achieved in three databases,with a weighted PLCC of 0.917 and SROCC of 0.912.According to the comprehensive benefits of accuracy,generalization and complexity,the IQA results were compared with those of 7 existing IQA models.The results showed that the accuracy of ICBS was 10.77%higher than those of SSIM in three databases,averagely,and its performance was better than that of 7 existing IQA models.Combined with the experimental results and theoretical analysis,it is showed that ICBS is an effective and excellent IQA model.