A Hybrid Quality Metric for Non-Integer Image Interpolation.

Jinling Chen,Yiwen Xu,Kede Ma,Huiwen Huang,Tiesong Zhao

DOI: https://doi.org/10.1109/qomex.2018.8463405

2018-01-01

Abstract:A great need of High-Resolution (HR) images has boosted the development of interpolation techniques. However, it is still a challenging task to objectively evaluate the perceptual quality of interpolated images, especially when the interpolation factor is a non-integer. To address this issue, we propose a hybrid quality metric for non-integer image interpolation that combines both reduced-reference and no-reference philosophies. To validate the proposed metric, we construct a non-integer interpolated image database and conduct a subjective user study to collect subjective opinions for each image. Experiments on the new database show that the proposed metric outperforms previous methods by a large margin.