A full reference quality assessment approach for screen content images based on high order derivative variation model

Abstract

In this work, we design a novel full reference (FR) quality evaluation approach of screen content images (SCIs) based on high order derivative variation model. The major contribution of this paper is the consideration that the human visual system (HVS) is sensitive to derivative information, and we apply the sensitivity property to evaluate the perceptual visual quality of SCIs. Specifically, we employ first-order derivative information to calculate quality map which quantifies the degradation of SCIs. Then, second-order derivative information is utilized to generate the weighting map. Finally, we get the overall quality score by incorporating the weighting map and quality map. The comparison experiments on a public SCI database demonstrate that the proposed approach can obtain the higher accuracy than other relevant ones in visual quality prediction of SCIs.