一种基于最小可察觉失真的立体图像质量客观评价方法

立体图像质量是评价立体视频系统性能的有效途径,而如何利用人类视觉特性对立体图像质量进行有效的评价是目前的研究难点。本文通过分析最小可察觉失真(JND,just noticeable distortion)视觉感知模型,并结合反映图像结构信息的奇异值矢量,提出了一种基于JND的立体图像质量客观评价方法。评价方法由图像质量评价和深度感知评价两部分组成,首先提取反映图像质量和深度感知的特征信息作为立体图像特征信息,然后根据立体图像的不同失真类型情况对其特征进行融合,通过支持向量回归(SVR,support vector Regression)预测得出立体图像质量的客观评价值。实验结果表明,采用本文提出的客观评价方法对立体数据测试库进行评价,在不同失真类型或混合失真评价结果中,Pearson线性相关系数(CC)值均在0.94以上,Spearman等级相关系数(SROCC)值均在0.92以上,符合人眼视觉特性,能够很好地预测人眼对立体图像的主观感知。

An objective quality assessment metric for stereoscopic images based on just noticeable distortion

In this paper,combining with the just noticeable distortion(JND) model and features of singular values,an objective quality assessment metric for stereoscopic images based on just noticeable distortion is proposed.The proposed metric consists of image quality assessment and depth perception assessment.Firstly,stereoscopic features are obtained by extracting the features of image quality and depth perception.Then,the features are fused according to different types of distortions.Finally,the values of objective assessment are predicted by support vector regression(SVR).Experimental results show that by applying the proposed model to stereoscopic test database,for either distortion quality assessment or mixture distortion quality assessment,the pearson linear correlation coefficient(CC) index reaches 0.94,and the spearman rank order correlation coefficient(SROCC) index reaches 0.92,which indicates that the model is fairly good and can predict human visual perception well.