EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Perceptual no-reference image quality assessment with meta-learning by graph representation learning and multi-scale feature fusion

Yanru Jia, Longsheng Wei iD

PLOS One · 18.06.2026 · 21 (6) · e0351549

DOI: 10.1371/journal.pone.0351549 ↗

Kurz gefasst

Fachartikel in der Zeitschrift PLOS One, online erschienen am 18.06.2026. Verfasst von Yanru Jia und 1 weiteren Autor:innen. Beteiligte Einrichtung: School of Computer Science and Artificial Intelligence, Xinyang College, Xinyang, Henan, China. Hauptthemen laut MEDLINE: Image Processing, Computer-Assisted, Visual Perception. Die Arbeit nennt 50 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

No-reference image quality assessment (NR-IQA) aims to predict perceptual quality in alignment with the human visual system (HVS), yet existing methods face challenges in capturing long-range dependencies across distortion types and levels while preserving content fidelity during preprocessing. This paper presents a perceptually-driven NR-IQA framework that integrates meta-learning, graph representation learning, and multi-scale feature fusion to address these limitations. First, a meta-learning paradigm is employed to pre-train a self-calibrated convolutional backbone, which adaptively models spatial and channel-wise dependencies across scales, thereby enhancing the extraction of distortion-aware features while mitigating information loss caused by fixed-input preprocessing. Second, a graph representation learning module is introduced to explicitly encode the hierarchical relationships among distortion types, distortion levels, and image content. Nodes in the graph correspond to distorted images, while edges capture inter-sample similarities; these are jointly optimized via a graph convolutional network under dual supervision from a triplet-based distortion-type discriminator and a probabilistic distortion-level regressor that accounts for content-induced uncertainty. Extensive experiments on four benchmark datasets demonstrate that our method achieves better performance, with average SROCC and PLCC improvements of 3.6–36.6% over hand-crafted feature-based methods and consistent gains over deep learning-based approaches. Ablation studies and visualizations confirm that the proposed components collectively yield a more discriminative and generalizable distortion representation, closely mirroring human perceptual judgments.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Image Processing, Computer-AssistedVisual PerceptionHumansRepresentation Machine LearningGraph Neural NetworksAlgorithmsConvolutional Neural Networks

Autor:innen und Institutionen

  • School of Computer Science and Artificial Intelligence, Xinyang College, Xinyang, Henan, China.
  • School of Artificial Intelligence and Automation, China University of Geosciences, Wuhan, China.

ORCID-Profile: Longsheng Wei

Zitieren

APA

Jia, Y., & Wei, L. (2026). Perceptual no-reference image quality assessment with meta-learning by graph representation learning and multi-scale feature fusion. PLOS One, 21(6), e0351549. https://doi.org/10.1371/journal.pone.0351549

Vancouver

Jia Y, Wei L. Perceptual no-reference image quality assessment with meta-learning by graph representation learning and multi-scale feature fusion. PLOS One. 2026;21(6):e0351549. doi:10.1371/journal.pone.0351549

BibTeX
@article{Jia2026,
  author = {Jia, Yanru and Wei, Longsheng},
  title = {Perceptual no-reference image quality assessment with meta-learning by graph representation learning and multi-scale feature fusion},
  journal = {PLOS One},
  year = {2026},
  volume = {21},
  number = {6},
  pages = {e0351549},
  doi = {10.1371/journal.pone.0351549}
}
Kostenlos starten →