A Soft-ranked Index Fusion Framework with Saliency Weighting for Image Quality Assessment
Liangwei Yu, Zhao Wang, Ye Yan, Lingyu Zhu, Shiqi Wang · 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · 2022
The compression technique is widely adopted for efficient data storage and transmission. Accurate image quality assessment (IQA) measures are urgently desired to evaluate the compression performance. To obtain a more robust evaluation, we propose a soft-ranked index fusion framework for the perceptual preference prediction task, with a combination of different quality measures. The derived soft-ranked indices are fully leveraged to provide the strong discriminability of ranking information. Furthermore, a saliency weighting approach is utilized to investigate the impact of visual attention on our framework. Experimental results indicate that our method achieves a promising prediction accuracy compared with the state-of-the-art quality measures.