基于大脑层状皮质模型的立体图像质量评价

Stereoscopic image quality assessment based on laminar cortical model

通过模拟人脑视觉神经接收视觉信息形成表面感知的处理机制,提出一种基于大脑层状皮质模型的全参考立体图像的图像质量评价(IQA)方法。首先,分析大脑形成表面感知的过程,提出可运用于立体图像的IQA的层状皮质模型;然后依据模型得到各层的响应输出,构建感知特征向量;最后利用机器学习算法,建立特征和质量的关系模型,预测立体图像质量。实验结果表明,本文方法在对称立体图像库上的Pearson线性相关系数(PLCC)和Spearman等级系数(SROCC)高于0.91,在非对称库上高于0.93。与现有的相关方法相比,本文方法与主观评价更加吻合,更适合立体图像的评价和优化。

By simulating the human visual nerve′ s processing mechanism of receiving visual information to form surface perceptio n,this paper proposes a full-reference stereoscopic image quality assessment （I QA） method based on the laminar cortical model.In the proposed method,we firstly analyze the process of forming the surface perception in the human brain,and propose a laminar cortical model for quality prediction.Then,fea tures are extracted by getting the response output of each layer based on the model.Finally,a regression model is learned to map human subjective scores and features by machine learning algorithm to predict the qual ity of a stereoscopic image. Based on the model,experimental results on three public databases demonstrate t hat the Pearson linear correlation coefficient （PLCC） and Spearman rank order correlation c oefficient （SROCC） of the proposed method are higher than 0.91in symmetric database and are higher than 0.93in asymmetric databases .The proposed algorithm achieves higher agreement with subjective assessment compared with the most related existing methods,making it better suited for the evaluation and optimization of stereoscopic image.