A Metric For Quantifying Image Quality Induced Saliency Variation

Pengfei Guo, Xin Zhao, Delu Zeng, Hantao LiuICIP 2021 · OpenReview

Abstract

Saliency plays an important role in the area of image quality assessment. Image distortions cause shift/redistribution of saliency from its original places. There is a need to be able to measure such distortion-included saliency variation (DSV), so that the use of saliency can be optimised for automated image quality assessment. Effort has been made in our previous study to build a benchmark for the measurement of DSV through subjective testing. In this paper, we demonstrate that exiting similarity measures are unhelpful for the quantification of DSV. Thus, we propose a new metric for DSV combining local and global measures using convex optimization. The experimental results show that our proposed metric can accurately quantify saliency variation.
BibTeX
@inproceedings{guo2021metric,
  author = {Pengfei Guo and Xin Zhao and Delu Zeng and Hantao Liu},
  title = {{A Metric For Quantifying Image Quality Induced Saliency Variation}},
  booktitle = {ICIP},
  year = {2021},
  url = {https://openreview.net/forum?id=rpweboAKtG8},
}

Loading related content...

Figure