Image Retargeting Quality Assessment using Structural Similarity and Information Preservation Rate

In order to display images appropriately in various devices with heterogeneous aspect ratios and sizes, there are different retargeting methods. These methods can be evaluated by objective measures more quickly and economically in comparison with subjective tests. Moreover, retargeting methods can utilize objective quality assessment algorithms to automatically adjust the image size considering its visual quality. In this paper, a new objective method is proposed to evaluate the visual quality of a retargeted image. The proposed scheme considers local structural similarity in conjunction with global information preserved after retargeting. The agreement of the proposed method with human visual system is analyzed over a publicly available CUHK retargeting dataset that contains subjective scores for each retargeted image. The experimental results demonstrate that the proposed method outperforms lots of the current methods and also have comparable accuracy in comparison with current methods which are remarkably more complex.