Objective Image Quality Assessment with Singular Value Decomposition

In this paper, we develop a new image quality assessment scheme with the use of Singular Value Decomposition (SVD). In a previous work [1], only the singular values out of SVD were used for quality assessment. We show that singular vectors are even more important than singular values for quality assessment. Hence, we use both singular vectors and values for comprehensive image quality evaluation, and demonstrate that the quality prediction is significantly improved. Extensive experimental results are reported as evidence of the efficiency of the proposed scheme, in comparison with the relevant existing methods, for three open and independent datasets with totally 2647 distorted images.