一种图像感知质量评价模型

A Perception Based Image Quality Assessment Model

在观察者观看和评价图像质量的过程中,提取对主观质量评价有重要影响的信息,并构建了一种图像感知质量评价模型.首先在主观评价过程中提取感知亮度、频率以及边缘3种重要信息;然后构造人眼视觉系统对3种信息的响应函数,并通过线性回归分析将函数拟合成为所需的图像感知质量评价模型;最后,分析了模型的预测值与主观质量评价方法给出的质量分数之间的Pearson相关系数和Spearman等级相关系数,分别为0.966和0.949.与其他一些有代表性的客观图像质量评价方法相比较,该模型具有明显的优越性.

Three important pieces of information are drawn out from the process of subjective image quality estimation. For each of the information, a model is established and also a function is found out to simulate its work on the human visual system. Then a multi-linear regression analysis is used to integrate these functions into an equation and used as our perception based image quality assessment model. Experimental results show that our estimation closely approximates human subjective tests such as mean opinion score, with high Pearson and Spearman correlation coefficients of 0.966 and 0.949. It is a significant improvement over some typical objective image quality estimations such as peak signal-to-noise ratio.