基于初级视皮层视觉特性的图像质量评价方法

Image Quality Assessment Based on Visual Characteristics of Primary Visual Cortex

针对现有的图像质量评价方法多从特征提取角度考虑图像的失真,忽视初级视皮层是视觉信息处理和认知推理的前提的问题,受人类视觉系统特性的启发,提出一种基于初级视皮层视觉特性的图像质量评价算法;该方法基于对初级视皮层视觉特性的学习,利用初级视皮层中感受野对视觉感知信息的稀疏编码特性,提取模拟初级视皮层感受野特性的基函数,结合独立成分分析和结构相似度算法构建一种初级视觉相似性测度法,并在LIVE图像数据库中进行实验。结果表明,该模型的预测结果与主观质量评价有很好的一致性,并优于已有的结构相似度算法。

In view that the existing image quality assessment methods mostly considered the image distortion from the perspective of feature extraction,ignoring that the primary visual cortex was the premise of visual information processing and cognitive reasoning,inspired by the characteristics of human visual system,an image quality assessment algorithm based on the visual features of primary visual cortex was proposed.The method was based on the study on the visual features of visual cortex,using the primary visual cortex receptive field in sparse coding characteristics of visual perception of information,extracted the basic functions of simulating the primary visual cortex receptive field characteristics.A new method primary visual similarity estimator(PRIVISE)for image quality assessment by combining independent components analysis with structural similarity index(SSIM)was built,then it was carried on the experiment in the LIVE image database.The results show that the model has good consistency with the image quality assessment and is better than the existing structural similarity algorithm.