TKScore for generated Thangka image quality assessment with perceptual alignment and structural constraints
Source abstract
Abstract
With rapid advances in generative models, generated image quality assessment has shifted from distribution-level evaluation toward instance-level and human-aligned assessment. However, existing metrics remain limited for generated Thangka images, whose quality depends on strict iconographic structures, symbolic color semantics, and fine decorative details. To address this challenge, we propose TKScore, a perception-aligned and structure-aware framework for generated Thangka image quality assessment. TKScore integrates adaptive region-weighted perceptual modeling for key iconographic regions, spatial-frequency collaborative feature extraction for texture and structural representation, and multi-dimensional attention regression for predicting color, texture, clarity, contour, and structure quality. Experiments on a self-built generated Thangka dataset show that TKScore outperforms representative traditional and modern baselines, achieving SRCC, PLCC, and KRCC values of 0.699, 0.728, and 0.516, respectively. These results demonstrate the effectiveness of TKScore for interpretable and human-aligned quality assessment of generated Thangka images.
Source: openalex_inverted_index · Quality: baseline
Research context
OpenAlex topics: Image and Video Quality Assessment · Visual Attention and Saliency Detection · Generative Adversarial Networks and Image Synthesis
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
References (35)
Show source-provided references
1. International Telecommunication Union. Methodologies for the subjective assessment of the quality of television images. Recommendation ITU-R BT.500-15 (International Telecommunication Union, 2023).
2. Likert, R. A technique for the measurement of attitudes. Arch. Psychol. 22, 1–55 (1932).
3. Liu, Q. et al. Image generation evaluation: a comprehensive survey of human and automatic evaluations. Front. Inf. Technol. Electron. Eng. 26, 1027–1065 (2025).10.1631/FITEE.2400904 ↗
4. Wang, Z., Bovik, A. C., Sheikh, H. R. & Simoncelli, E. P. Image quality assessment: from error visibility to structural similarity. IEEE Trans. Image Process. 13, 600–612 (2004).10.1109/TIP.2003.819861 ↗
5. Zhang, R., Isola, P., Efros, A. A., Shechtman, E. & Wang, O. The unreasonable effectiveness of deep features as a perceptual metric. In Proc. IEEE conference on computer vision and pattern Recognition (CVPR) 586–595 (IEEE, 2018).10.1109/CVPR.2018.00068 ↗
6. Salimans, T. et al. Improved techniques for training GANs. Adv. Neural Inform. Process. Syst. 29, 2234–2242 (2016).
7. Heusel, M. et al. GANs trained by a two time-scale update rule converge to a local Nash equilibrium. Adv. Neural Inform. Process. Syst. 30, 6626–6637 (2017).
8. Bińkowski, M., Sutherland, D. J., Arbel, M. & Gretton, A. Demystifying MMD GANs. In International Conference on Learning Representations (ICLR), 1–36 (OpenReview.net, 2018).
9. Bai, C.-Y., Lin, H.-T., Raffel, C. & Kan, W.-C. On training sample memorization: lessons from benchmarking generative modeling with a large-scale competition. In Proc. 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining, 2534–2542 (ACM, 2021).10.1145/3447548.3467198 ↗
10. Lopez-Paz, D. & Oquab, M. Revisiting classifier two-sample tests. In International Conference on Learning Representations (ICLR), 1–15 (OpenReview.net, 2017).
11. Sajjadi, M. S. M., Bachem, O., Lucic, M., Bousquet, O. & Gelly, S. Assessing generative models via precision and recall. Adv. Neural Inform. Process. Syst. 31, 5228–5237 (2018).
12. Kynkäänniemi, T., Karras, T., Laine, S., Lehtinen, J. & Aila, T. Improved precision and recall metric for assessing generative models. Adv. Neural Inform. Process. Syst. 32, 3927-3936 (2019).
13. Gu, S., Bao, J., Chen, D. & Wen, F. GIQA: generated image quality assessment. in Computer Vision – ECCV 2020 Workshops 369–385 (Springer, 2020).10.1007/978-3-030-58621-8_22 ↗
14. Yang, Y. & Fan, F. Ancient thangka Buddha face recognition based on the Dlib machine learning library and comparison with secular aesthetics. Herit. Sci. 11, 137 (2023).10.1186/s40494-023-00983-8 ↗
15. Xian, Y., Xiang, Y., Yang, X., Zhao, Q. & Cairang, X. Thangka school image retrieval based on multi-attribute features. npj Herit. Sci. 13, 179 (2025).10.1038/s40494-025-01762-3 ↗
16. Yang, Y. et al. Learning multi-granularity features for re-identifying figures in portrait Thangka images. In Proc. 26th International Conference on Pattern Recognition (ICPR), 1643–1649 (IEEE, 2022).10.1109/ICPR56361.2022.9956311 ↗
17. Chi, Z. et al. Interactive Thangka image restoration guided by line art. J. Image Graph. 30, 1377–1388 (2025).10.11834/jig.240399 ↗
18. Chen, X. et al. Thangka super-resolution diffusion model based on discrete cosine transform domain padding upsampling and high-frequency focused attention. PLoS ONE 20, e0332904 (2025).10.1371/journal.pone.0332904 ↗
19. Liu, A., Lin, W., Paul, M., Deng, C. & Zhang, F. Just noticeable difference for images with decomposition model for separating edge and textured regions. IEEE Trans. Circuits Syst. Video Technol. 20, 1648–1652 (2010).10.1109/TCSVT.2010.2087432 ↗
20. Hu, J., Shen, L. & Sun, G. Squeeze-and-excitation networks. In Proc. IEEE/CVF Conference On Computer Vision And Pattern Recognition (CVPR), 7132–7141 (IEEE, 2018).10.1109/CVPR.2018.00745 ↗
21. Vaswani, A. et al. Attention is all you need. Adv. Neural Inform. Process. Syst. 30, 5998–6008 (2017).
22. Su, S. et al. Blindly assess image quality in the wild guided by a self-adaptive hyper network. In Proc. IEEE/CVF Conference On Computer Vision And Pattern Recognition (CVPR), 3664–3673 (IEEE, 2020).10.1109/CVPR42600.2020.00372 ↗
23. Tang, Z., Wang, Z., Peng, B. & Dong, J. CLIP-AGIQA: boosting the performance of AI-generated image quality assessment with CLIP. In Proc. 27th International Conference on Pattern Recognition (ICPR), 48–61 (Springer, 2025).10.1007/978-3-031-78125-4_4 ↗
24. Wu, H. et al. Q-Align: teaching LMMs for visual scoring via discrete text-defined levels. In Proc. 41st International Conference on Machine Learning, Vol. 235, 54015–54029 (PMLR, 2024).
25. Agnolucci, L., Galteri, L. & Bertini, M. Quality-aware image-text alignment for opinion-unaware image quality assessment. Preprint at https://doi.org/10.48550/arXiv.2403.11176 (2024).10.48550/arXiv.2403.11176 ↗
26. Yuan, J. et al. PKU-I2IQA: an image-to-image quality assessment database for AI-generated images. Preprint at https://doi.org/10.48550/arXiv.2311.15556 (2023).10.48550/arXiv.2311.15556 ↗
27. Pressman, J. D., Crowson, K. & Simulacra Captions Contributors. Simulacra Aesthetic Captions, Version 1.0. GitHub https://github.com/JD-P/simulacra-aesthetic-captions (2022).
28. Qin, X. et al. U2-Net: going deeper with nested U-structure for salient object detection. Pattern Recognit. 106, 107404 (2020).10.1016/j.patcog.2020.107404 ↗
29. Hou, Q. et al. Deeply supervised salient object detection with short connections. In Proc. IEEE/CVF Conference On Computer Vision And Pattern Recognition (CVPR) 5300–5309 (IEEE, 2017).10.1109/CVPR.2017.563 ↗
30. Qin, X. et al. BASNet: boundary-aware salient object detection. In Proc. IEEE/CVF Conference On Computer Vision And Pattern Recognition (CVPR), 7471–7481 (IEEE, 2019).10.1109/CVPR.2019.00766 ↗
31. Zhao, J. et al. EGNet: edge guidance network for salient object detection. In Proc. IEEE/CVF International Conference on Computer Vision (ICCV), 8778–8787 (IEEE, 2019).10.1109/ICCV.2019.00887 ↗
32. He, K., Zhang, X., Ren, S. & Sun, J. Deep residual learning for image recognition. In Proc. IEEE/CVF Conference On Computer Vision And Pattern Recognition (CVPR), 770–778 (IEEE, 2016).10.1109/CVPR.2016.90 ↗
33. Tan, M. & Le, Q. V. EfficientNet: rethinking model scaling for convolutional neural networks. In Proc. 36th International Conference on Machine Learning (ICML), Proceedings of Machine Learning Research, Vol. 97, 6105–6114 (PMLR, 2019).
34. Liu, Z. et al. A ConvNet for the 2020s. In Proc. IEEE/CVF Conference On Computer Vision And Pattern Recognition (CVPR), 11966–11976 (IEEE, 2022).10.1109/CVPR52688.2022.01167 ↗
35. Dosovitskiy, A. et al. An image is worth 16 × 16 words: transformers for image recognition at scale. In International Conference on Learning Representations (ICLR), 1–22 (OpenReview.net, 2021).