中文摘要：

提出了一种基于稀疏表示的立体图像质量评价方法，分为训练和测试两个部分。在训练部分，通过训练不同频带的立体图像获得立体图像的稀疏字典；在测试部分，根据稀疏字典计算得到立体图像的稀疏特征，定义了稀疏特征相似度衡量原始和失真图像信息的差异，并根据稀疏字典计算了频带增益和左右视点的融合权值，最后融合稀疏特征相似度作为立体图像质量的客观评价值。在立体图像测试库上的实验结果表明，本文方法的评价结果与主观评价结果有较好的相关性，符合人类视觉系统的感知。

英文摘要：

Stereoscopic image quality assessment is an effective way to evaluate the performance of stereo-scopic video system,and objective stereoscopic image quality assessment consistent with subjective per- ception is still a great challenge in image quality assessment. In this paper, an objective quality assess- ment method for stereoscopic image based on sparse representation is proposed. The proposed method is composed of two stages, training and testing. In the training stage, sparse dictionaries on different fre- quency channels are learned from the training images. In the testing stage, sparse features for all images are extracted based on the learnt sparse dictionaries, and sparse feature similarities between the reference and distorted images are calculated for each view. In addition, the gain of each channel and weights of two views are computed to model the binocular physiological behaviors. Finally, by fusing the gain and weights,the objective quality score is obtained. Experimental results show that the proposed metric can achieve better consistency with subjective assessment, which indicates that the metric can predict human visual perception very well.