Preselection Based Subjective Preference Evaluation for the Quality of Underwater Images

Miao Yang, Yixiang Du, Yue Huang, Hantao Liu, Zhiqiang Wei, Jintong Hu, Ke Hu, Zhibin Sheng; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2019, pp. 34-43

Underwater images contain an interactive mixture of distortions due to the physicochemical property of water and the instability of imaging systems, which differ from those in natural images. We cannot obtain the pristine underwater image as the reference applied in the traditional benchmark databases, and the groups of gradual distortions either. In this paper, a novel preselection based preference label evaluation method is proposed to construct a combined subjective test procedure for an extended preference judgment dataset of underwater images. To the best of our knowledge, this is the first subjective evaluation procedure for underwater images, and also a solution for an expanding visual preference benchmark database. We demonstrate the excellent correlation of the proposed subjective evaluation with the traditional image quality assessment. It is also proven that the proposed subjective evaluation procedure could reflect the slight change of image quality and the authentic quality of a picture more accurately better than the traditional methods.