作　　者： ; ; ; ; ;

机构地区： 宁波大学信息科学与工程学院

出　　处： 《光电子．激光》 2014年第1期192-198,共7页

摘　　要： 根据梯度结构张量能够表示图像结构信息的特点,提出了一种基于三维结构张量的立体图像客观质量评价方法。首先分别求取原始和失真的立体图像水平、垂直和视点方向的梯度信息,以及敏感区域,并构造出立体图像中每个像素的三维结构张量矩阵;然后,提取三维结构张量矩阵的特征值和特征向量信息;最后,根据特征值和特征向量预测得到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图像测试库进行评价,总体评价的Pearson线性相关系数(PLCC)和Spearman等级相关系数(SROCC)值均在0.92左右,Kendall相关系数(KROCC)值接近0.80,均方根误差(RMSE)值均在6.00左右;与其他方法相比,本方法具有较高的预测精确性。 Stereoscopic image quality assessment is an effective way to evaluate the performance of stereoscopic video system.However,how to utilize human visual characteristics in quality assessme nt is still an issue.In this paper,considering the characteristics of gradient s tructure tensor in image feature description,an objective stereoscopic image quality assessment method based on three-dimensional （3D） structure tensor is p roposed.To be more specific,we obtain the horizontal,vertical and inter-view gradient information for each pixel in the original and distorted stereoscopic images,and construct the 3D structure te nsor.Then,the corresponding eigenvalues and eigenvectors are extracted from the 3D structure tensor.Finally,according to the eigenvalues and the eigenvectors,th e values of objective assessment between the original and distorted stereoscopic images are predicted.Experimental results d emonstrate that compared with other methods,the overall Pearson linear correlation coefficient （PLCC） and the Spearman rank order correlation coefficient （SROCC） indicators reach 0.92,the Kendall rank-order correlation coefficient （KROCC） indicator reaches 0. 80,and the root mean squared error （RMSE） indicator is approxim ately 6.00,w hich indicates that the proposed method can achieve higher prediction accuracy.

关 键 词： 立体图像质量评价 三维结构张量 梯度信息

领　　域： [电子电信] [电子电信]