\r李素梅，常永莉，韩旭，胡佳洁\r
\r
AuthorsHTML:\r李素梅，常永莉，韩旭，胡佳洁\r
\r
AuthorsListE:\rLi Sumei，Chang Yongli，Han Xu，Hu Jiajie\r
\r
AuthorsHTMLE:\rLi Sumei，Chang Yongli，Han Xu，Hu Jiajie\r
\r
Unit:\r天津大学电气自动化与信息工程学院，天津 300072\r
\r
Unit_EngLish:\rSchool of Electrical and Information Engineering，Tianjin University，Tianjin 300072，China\r
\r
Abstract_Chinese:\r本文提出了一种基于稀疏字典学习的双通道立体图像质量评价方法．其中，一个通道结合视觉注意机制得到初始立体显著图，用中央偏移和中心凹特性对其进行优化得到最终的显著图，然后，对其进行稀疏字典训练获得显著字典；另一个通道将参考立体图像对进行SIFT 特征变换，然后，对其进行稀疏字典训练获得SIFT 字典．在测试阶段，利用已训练字典对参考图像和失真图像进行稀疏编码获得稀疏系数，并定义稀疏系数相似度指标以衡量参考图像和失真图像之间的信息差异；最后将两个通道的质量分数进行加权得到立体图像质量的客观分数．实验在两个公开LIVE 库上进行测试，实验结果表明，本文算法的评价结果与主观评分具有更好的一致性，更加符合人类视觉系统的感知．\r
\r
Abstract_English:\rIn this paper，a dual-channel quality assessment method of stereoscopic images using sparse representation was proposed. For one channel，the initial 3D salient map was obtained with visual attention mechanism，and the final salient map was generated through optimization by centre bias and the fovea property，then is used to train a salient dictionary. For another channel，a scale-invariant feature transform(SIFT) dictionary was trained by SIFT features extracted from reference images. At the testing stage，the trained dictionaries were used to get the sparse coefficient matrices for reference images and distorted images，and a sparse coefficient similarity index was defined to measure the information difference between them. Finally，the quality scores of the two channels were pooled to achieve the object score of the stereoscopic image. The performance of the proposed stereoscopic image quality evaluation metric was verified on two publicly available LIVE databases，and experimental results show that the proposed algorithm achieves high consistent alignment with subjective assessment，which satisfies the HVS better.\r
\r
Keyword_Chinese:立体图像质量评价；稀疏字典；视觉显著性；SIFT 特征；中央偏移；中心凹\r

Keywords_English:stereoscopic image quality evaluation；sparse dictionary；visual saliency；SIFT feature；center bias(CB)；fovea\r


PDF全文下载地址:http://xbzrb.tju.edu.cn/#/digest?ArticleID=6170