Dong, Zhengyan, Liu, Xiaochang, Wu, Xinbo, Wu, Yingying, Liu, Jiang, Ma, Yueran, Chen, Ying and Liu, Hantao
2026.
Study of emotion in image quality assessment.
IEEE Transactions on Multimedia
10.1109/TMM.2026.3716095
|
Preview |
PDF
- Accepted Post-Print Version
Available under License Creative Commons Attribution. Download (12MB) | Preview |
Abstract
Traditional image quality assessment (IQA) methods primarily focus on lower-level attributes such as sharpness, noise, and compression artefacts. However, as visual content increasingly serves communicative and expressive purposes, the emotional impact of images has become a critical factor in evaluating perceived quality. To systematically study emotion in IQA, we adopt both subjective and objective methodologies. First, we conduct a fully controlled psychovisual experiment in which human subjects rate the quality of 600 emotion-rich images of various distortions. This study yields CUeIQA, a novel IQA database specifically designed to capture the interplay between emotional content and perceived quality. Building upon this, we propose eIQANet, a deep learning-based IQA model that integrates an emotion-cognisant module to capture diverse emotional effects and combines these with distortion-based quality features. Experimental results demonstrate eIQANet outperforms existing methods in assessing perceived quality of emotion-rich images. This work advances IQA by incorporating emotional perception into quality assessment, addressing a key limitation of conventional methodologies.
| Item Type: | Article |
|---|---|
| Date Type: | Published Online |
| Status: | In Press |
| Schools: | Schools > Computer Science & Informatics |
| Publisher: | Institute of Electrical and Electronics Engineers |
| ISSN: | 1520-9210 |
| Date of First Compliant Deposit: | 23 July 2026 |
| Date of Acceptance: | 12 April 2026 |
| Last Modified: | 25 Aug 2026 22:45 |
| URI: | https://orca.cardiff.ac.uk/id/eprint/188433 |
Actions (repository staff only)
 |
Edit Item |
Dimensions
Dimensions