SSVD: Structural SVD-based image quality assessment

Abstract

In the last decade, some impressive image quality metrics have been proposed; however, designing an image quality metric which predicts human judgments is still a challenging issue. It is due to the complexity of the human visual system. Singular value decomposition (SVD), as a useful tool, has been employed for evaluating the perceptual quality of visual information. The efficiency of the SVD-based image quality assessment (IQA) methods is related to its ability to extract the structural information of the viewing scene. In this paper, a new SVD-based IQA method is presented in which the structural information of the distorted image is evaluated based on its reflection on the original singular vector matrices. The experimental results show that the proposed algorithm can effectively evaluate the natural image quality in a consistent manner with the human visual perception.