Full search page →

SUPIR: scaling up the prior, the data and the text (CVPR 2024)

Optional · paper guide 11 of 11 for DLCV 09A.1 · Blind and Real-World Image Super-Resolution. Citation numbers refer to the reference list at the bottom of this page.

Navigate: this page and this track: Navigate: this page and this track

On this page

Supplementary notes · 3

  1. The high-order degradation pipeline, step by step
  2. Sinc filters and ringing
  3. Why the kernel is hard to identify

Paper guides · 11

  1. SRMD: one network for many known degradations (CVPR 2018)
  2. IKC: iterative kernel correction (CVPR 2019)
  3. KernelGAN: kernel estimation with an internal GAN (NeurIPS 2019)
  4. DAN: unfolding the alternating optimization (NeurIPS 2020)
  5. DASR: contrastive degradation representations (CVPR 2021)
  6. BSRGAN: a practical degradation model (ICCV 2021)
  7. Real-ESRGAN: high-order degradation with pure synthetic data (ICCVW 2021)
  8. StableSR: a time-aware encoder for a frozen diffusion prior (IJCV 2024)
  9. DiffBIR: remove the degradation, then regenerate (ECCV 2024)
  10. SeeSR: degradation-aware semantic prompts (CVPR 2024)
  11. SUPIR: scaling up the prior, the data and the text (CVPR 2024)

Deep Learning for Computer Vision

  1. 02 · Deep Image Classification
View the whole course tree →

Challenge

A larger generative prior can provide more convincing texture, but restoration needs control over the observed image as well as visual realism. SUPIR investigates model scaling, a much larger restoration-training collection and language guidance together [30]. Simply attaching a standard controller to a powerful text-to-image generator can produce attractive details that drift from the input.

Three difficulties motivate its components: the ordinary autoencoder can misread degraded images, a large controller is expensive, and aggressive generative sampling can invent structure. Its answer is a degradation-robust encoder, a trimmed controller with stronger feature modulation, and restoration-guided sampling. The system targets perceptual restoration; exact recovery of the original high-frequency scene is not assured.

Method

Compare the copied/trimmed encoder route and ZeroSFT connector; low-quality image features guide the frozen large prior through an explicitly trained adapter.
Original paper figure 3 — Compare the copied/trimmed encoder route and ZeroSFT connector; low-quality image features guide the frozen large prior through an explicitly trained adapter.

Figure source: [30], arXiv:2401.13627v2, PDF page 4. Original PDF.

Pretrained prior. The backbone is SDXL [80], kept frozen during adaptation. The paper discusses a 2.6-billion-parameter generative prior; this should not be read as the complete deployment footprint, which also includes conditioning, image encoding and text processing. Larger prior capacity is useful only if image guidance can keep it tied to the particular observation.

Degradation-robust encoding. Starting from the pretrained encoder, train an encoder EdrE_{\mathrm{dr}} so that a fixed decoder DD produces similar reconstructions from degraded and clean partners:

LE=∥D(Edr(y))−D(Edr(x))∥22.\mathcal L_E=\|D(E_{\mathrm{dr}}(y))-D(E_{\mathrm{dr}}(x))\|_2^2.

Here xx is an HR training image and yy its degraded, enlarged partner. Both are encoded into the prior’s latent coordinate system. The decoded comparison encourages damage invariance in the conditioning representation, rather than asking the diffusion network to decide whether a JPEG block is meaningful content. This does not imply that identical codes are possible for every ambiguous pair.

Trimmed controller and ZeroSFT. A copy of the generative encoder supplies a control path, but the authors remove half of the transformer blocks within its encoder blocks to reduce overhead. The ZeroSFT connector combines zero-initialized convolution, normalization and spatial feature transformation. It supplies stronger affine image guidance than a residual-only connector while preserving stable initialization. This is a distinction between the control-path architecture and the frozen generator; trimming the controller does not remove half the frozen prior.

Image–text training. The paper collects 20 million high-quality image–text examples and additional face data, then generates degraded inputs with a Real-ESRGAN-style pipeline [17]. Captioning helps describe the intended content. Negative-quality examples teach the model what undesirable low-quality appearance means. Without these training examples, a negative prompt alone is not guaranteed to have the intended restoration effect. At inference, a multimodal captioning model [81] can describe the processed input, and a user may edit the text. Such text is extra conditioning, not an independent observation of the photographed scene.

Restoration-guided sampling. The method modifies an EDM sampler [82]. At noise level σt\sigma_t, its denoised prediction is pulled toward the degraded-image latent zyz_y with a weight

kt=(σt/σT)τr,z~=z^+kt(zy−z^).k_t=(\sigma_t/\sigma_T)^{\tau_r}, \qquad \widetilde z=\widehat z+k_t(z_y-\widehat z).

Here z^\widehat z is the denoised estimate, σT\sigma_T the starting noise scale, and τr\tau_r the restoration-control exponent. The sampler uses the corrected estimate to compute its update direction; this is not simply overwriting the final image. The pull is stronger at high noise levels and fades as detail forms. A smaller exponent maintains stronger input influence later. The reported default uses 100 sampling steps and τr=4\tau_r=4.

Results

Table 2a reports scores on 60 real low-quality images: SUPIR has CLIPIQA 0.8232, MUSIQ 73.00 and MANIQA 0.4295. StableSR in that comparison has 0.7654, 70.70 and 0.3035. The metrics predict preferred appearance; they cannot confirm that a generated texture is correct. On the synthetic ×4 task in Table 1, SUPIR gives PSNR 22.66, SSIM 0.5763 and LPIPS 0.2662, whereas BSRGAN gives 24.97, 0.6572 and 0.2261. Strong no-reference scores can coexist with worse reference fidelity.

These tables use the paper’s input and output processing. SUPIR’s larger-resolution operating point differs from small-crop benchmarks used by other papers. Do not combine scores from those protocols into a single ranking without rerunning the methods consistently.

Ablations

Table 2c replaces ZeroSFT with zero convolution. PSNR changes from 20.66 to 19.47, SSIM from 0.4763 to 0.4261 and LPIPS from 0.3412 to 0.3969. Yet MUSIQ rises from 73.66 to 74.00, again demonstrating that extra attractive detail can raise an NR score while harming fidelity.

Table 2b studies prompts and negative examples. With negative training samples but neither positive nor negative quality prompts, MUSIQ is 68.94. With both prompts it is 73.66, while PSNR changes from 22.90 to 20.66. Removing negative training examples but keeping both prompts gives MUSIQ 72.68 and MANIQA 0.3180, versus 0.4740 for the complete setting. Other visual ablations study model scale, degradation-robust encoding and the restoration exponent; their effects should be read under the associated training conditions, not attributed to scale alone.

Conclusion

Scaling a restoration prior is a system-design problem. Representation robustness, controller strength, image–text data and sampling constraints are all needed to make the larger generator useful for a specific degraded input.

Problems

The system is large and multi-step. Caption generation adds work and can supply incorrect semantics. Prompt edits may change image attributes rather than recover them, and strong guidance can retain input defects. Metric improvements may reflect aesthetics instead of correctness. Its value is perceptually convincing restoration under stated assumptions; applications that depend on factual detail need stronger measurement constraints and explicit uncertainty handling.

References

  1. N. Efrat, D. Glasner, A. Apartsin, B. Nadler and A. Levin, “Accurate Blur Models vs. Image Priors in Single Image Super-Resolution,” in Proc. IEEE ICCV, 2013, pp. 2832–2839. doi · CVF
  2. K. Zhang, W. Zuo and L. Zhang, “Learning a Single Convolutional Super-Resolution Network for Multiple Degradations,” CVPR, 2018. paper
  3. A. Liu, Y. Liu, J. Gu, Y. Qiao, C. Dong, “Blind Image Super-Resolution: A Survey and Beyond,” IEEE TPAMI (early access), 2022. source
  4. Honggang Chen, Xiaohai He, Linbo Qing, Yuanyuan Wu, Chao Ren, Ray E. Sheriff, Ce Zhu, “Real-world single image super-resolution: A brief review,” Information Fusion, 2022. source
  5. B. B. Moser, A. S. Shanbhag, F. Raue, S. Frolov, S. Palacio and A. Dengel, “Diffusion Models, Image Super-Resolution, and Everything: A Survey,” IEEE Transactions on Neural Networks and Learning Systems, vol. 36, no. 7, pp. 11793–11813, 2025. doi · arXiv
  6. Y. Xu, S. R. Tseng, Y. Tseng, H. Kuo, Y. Tsai, “Unified Dynamic Convolutional Network for Super-Resolution with Variational Degradations,” CVPR, 2020. source
  7. K. Zhang, L. Van Gool, R. Timofte, “Deep Unfolding Network for Image Super-Resolution,” CVPR, 2020. source
  8. J. Gu, H. Lu, W. Zuo and C. Dong, “Blind Super-Resolution With Iterative Kernel Correction,” CVPR, 2019. paper
  9. Z. Luo, Y. Huang, S. Li, L. Wang and T. Tan, “Unfolding the Alternating Optimization for Blind Super Resolution,” NeurIPS, 2020. paper
  10. Z. Luo, H. Huang, L. Yu, Y. Li, H. Fan and S. Liu, “Deep Constrained Least Squares for Blind Image Super-Resolution,” CVPR, 2022. paper
  11. T. Michaeli, M. Irani, “Nonparametric Blind Super-resolution,” Proceedings of the IEEE International Conference on Computer Vision, 2013. source
  12. S. Bell-Kligler, A. Shocher and M. Irani, “Blind Super-Resolution Kernel Estimation using an Internal-GAN,” NeurIPS, 2019. paper
  13. A. Shocher, N. Cohen and M. Irani, “‘Zero-Shot’ Super-Resolution Using Deep Internal Learning,” in Proc. IEEE/CVF CVPR, 2018, pp. 3118–3126. doi · arXiv
  14. L. Wang, Y. Wang, X. Dong, Q. Xu, J. Yang, W. An and Y. Guo, “Unsupervised Degradation Representation Learning for Blind Super-Resolution,” CVPR, 2021. paper
  15. B. Xia, Y. Zhang, Y. Wang, Y. Tian, W. Yang, R. Timofte and L. Van Gool, “Knowledge Distillation Based Degradation Estimation for Blind Super-Resolution,” ICLR, 2023. paper
  16. K. Zhang, J. Liang, L. Van Gool, R. Timofte, “Designing a Practical Degradation Model for Deep Blind Image Super-Resolution,” Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021. source
  17. X. Wang, L. Xie, C. Dong, Y. Shan, “Real-ESRGAN: Training Real-World Blind Super-Resolution With Pure Synthetic Data,” Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021. source
  18. J. Liang, H. Zeng, L. Zhang, “Efficient and Degradation-Adaptive Network for Real-World Image Super-Resolution,” ECCV, 2022. source
  19. J. Cai, H. Zeng, H. Yong, Z. Cao, L. Zhang, “Toward Real-World Single Image Super-Resolution: A New Benchmark and a New Model,” Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019. source
  20. P. Wei, Z. Xie, H. Lu, Z. Zhan, Q. Ye, W. Zuo, L. Lin, “Component Divide-and-Conquer for Real-World Image Super-Resolution,” ECCV, 2020. source
  21. C. Chen, Z. Xiong, X. Tian, Z. Zha, F. Wu, “Camera Lens Super-Resolution,” Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019. source
  22. Y. Yuan, S. Liu, J. Zhang, Y. Zhang, C. Dong, L. Lin, “Unsupervised Image Super-Resolution using Cycle-in-Cycle Generative Adversarial Networks,” CVPR Workshops, 2018. source
  23. A. Bulat, J. Yang, G. Tzimiropoulos, “To learn image super-resolution, use a GAN to learn how to do image degradation first,” ECCV, 2018. source
  24. M. Fritsche, S. Gu, R. Timofte, “Frequency Separation for Real-World Super-Resolution,” ICCV Workshops, 2019. source
  25. X. Ji, Y. Cao, Y. Tai, C. Wang, J. Li, F. Huang, “Real-World Super-Resolution via Kernel Estimation and Noise Injection,” Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2020. source
  26. J. Wang, Z. Yue, S. Zhou, K. C. K. Chan, C. C. Loy, “Exploiting Diffusion Prior for Real-World Image Super-Resolution,” IJCV, 2024. source
  27. X. Lin, J. He, Z. Chen, et al., “DiffBIR: Towards Blind Image Restoration with Generative Diffusion Prior,” ECCV, 2024. source
  28. T. Yang, R. Wu, P. Ren, X. Xie, L. Zhang, “Pixel-Aware Stable Diffusion for Realistic Image Super-resolution and Personalized Stylization,” ECCV, 2024. source
  29. R. Wu, T. Yang, L. Sun, Z. Zhang, S. Li, L. Zhang, “SeeSR: Towards Semantics-Aware Real-World Image Super-Resolution,” Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024. source · supplementary experiments
  30. F. Yu, J. Gu, Z. Li, et al., “Scaling Up to Excellence: Practicing Model Scaling for Photo-Realistic Image Restoration In the Wild,” CVPR, 2024. source
  31. Z. Yue, J. Wang, C. C. Loy, “ResShift: Efficient Diffusion Model for Image Super-resolution by Residual Shifting,” NeurIPS, 2023. source
  32. X. Wang, K. Yu, C. Dong, C. C. Loy, “Recovering Realistic Texture in Image Super-Resolution by Deep Spatial Feature Transform,” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018. source
  33. K. He, H. Fan, Y. Wu, S. Xie, R. Girshick, “Momentum Contrast for Unsupervised Visual Representation Learning,” CVPR, 2020. source
  34. X. Wang, K. Yu, S. Wu, J. Gu, Y. Liu, C. Dong, Y. Qiao and C. C. Loy, “ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks,” in Proc. ECCV Workshops, 2018 (Lecture Notes in Computer Science, 2019), pp. 63–79. doi · arXiv
  35. O. Ronneberger, P. Fischer and T. Brox, “U-Net: Convolutional Networks for Biomedical Image Segmentation,” in Proc. MICCAI, 2015, pp. 234–241. doi · arXiv
  36. T. Miyato, T. Kataoka, M. Koyama, Y. Yoshida, “Spectral Normalization for Generative Adversarial Networks,” ICLR, 2018. source
  37. J. Liang, H. Zeng, L. Zhang, “Details or Artifacts: A Locally Discriminative Learning Approach to Realistic Image Super-Resolution,” Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022. source
  38. A. Ignatov, N. Kobyshev, R. Timofte, K. Vanhoey, L. Van Gool, “DSLR-Quality Photos on Mobile Devices with Deep Convolutional Networks,” ICCV, 2017. source
  39. A. Lugmayr, M. Danelljan, R. Timofte, “Unsupervised Learning for Real-World Super-Resolution,” arXiv preprint, 2019. source
  40. A. Lugmayr, M. Danelljan, R. Timofte, et al., “NTIRE 2020 Challenge on Real-World Image Super-Resolution: Methods and Results,” CVPR Workshops, 2020. source
  41. R. Rombach, A. Blattmann, D. Lorenz, P. Esser, B. Ommer, “High-Resolution Image Synthesis with Latent Diffusion Models,” CVPR, 2022. source
  42. L. Zhang, A. Rao, M. Agrawala, “Adding Conditional Control to Text-to-Image Diffusion Models,” ICCV, 2023. source
  43. Y. Ai, X. Zhou, H. Huang, X. Han, Z. Chen, Q. You, H. Yang, “DreamClear: High-Capacity Real-World Image Restoration with Privacy-Safe Dataset Curation,” NeurIPS, 2024. source
  44. Z. Duan, J. Zhang, X. Jin, et al., “DiT4SR: Taming Diffusion Transformer for Real-World Image Super-Resolution,” ICCV, 2025. source
  45. R. Wu, L. Sun, Z. Ma, L. Zhang, “One-Step Effective Diffusion Network for Real-World Image Super-Resolution,” NeurIPS, 2024. source
  46. Y. Wang, W. Yang, X. Chen, et al., “SinSR: Diffusion-Based Image Super-Resolution in a Single Step,” CVPR, 2024. source
  47. E. Agustsson and R. Timofte, “NTIRE 2017 Challenge on Single Image Super-Resolution: Dataset and Study,” in Proc. IEEE CVPR Workshops, 2017, pp. 1122–1131. doi · CVF
  48. Z. Wang, A. C. Bovik, H. R. Sheikh and E. P. Simoncelli, “Image Quality Assessment: From Error Visibility to Structural Similarity,” IEEE Transactions on Image Processing, vol. 13, no. 4, pp. 600–612, 2004. doi
  49. R. Zhang, P. Isola, A. A. Efros, E. Shechtman and O. Wang, “The Unreasonable Effectiveness of Deep Features as a Perceptual Metric,” in Proc. IEEE/CVF CVPR, 2018, pp. 586–595. doi · arXiv
  50. K. Ding, K. Ma, S. Wang, E. P. Simoncelli, “Image Quality Assessment: Unifying Structure and Texture Similarity,” IEEE TPAMI, 2022. source
  51. A. Mittal, R. Soundararajan and A. C. Bovik, “Making a “Completely Blind” Image Quality Analyzer,” IEEE Signal Processing Letters, vol. 20, no. 3, pp. 209–212, 2013. doi
  52. J. Ke, Q. Wang, Y. Wang, P. Milanfar, F. Yang, “MUSIQ: Multi-scale Image Quality Transformer,” ICCV, 2021. source
  53. S. Yang, T. Wu, S. Shi, S. Lao, Y. Gong, M. Cao, J. Wang, Y. Yang, “MANIQA: Multi-dimension Attention Network for No-Reference Image Quality Assessment,” CVPR Workshops, 2022. source
  54. J. Wang, K. C. K. Chan, C. C. Loy, “Exploring CLIP for Assessing the Look and Feel of Images,” AAAI, 2023. source
  55. M. Heusel, H. Ramsauer, T. Unterthiner, B. Nessler, S. Hochreiter, “GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium,” NeurIPS, 2017. source
  56. R. Timofte, S. Gu, J. Wu, L. Van Gool, “NTIRE 2018 Challenge on Single Image Super-Resolution: Methods and Results,” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, 2018. source
  57. J. Cai, S. Gu, R. Timofte, L. Zhang, “NTIRE 2019 Challenge on Real Image Super-Resolution: Methods and Results,” Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2019. source
  58. A. Lugmayr, M. Danelljan, R. Timofte, et al., “AIM 2019 Challenge on Real-World Image Super-Resolution: Methods and Results,” ICCV Workshops, 2019. source
  59. P. Wei, H. Lu, R. Timofte, et al., “AIM 2020 Challenge on Real Image Super-Resolution: Methods and Results,” ECCV Workshops, 2020. source
  60. Z. Chen, K. Liu, J. Gong, J. Wang, L. Sun, Z. Wu, R. Timofte, Y. Zhang et al., “NTIRE 2025 Challenge on Image Super-Resolution (×4): Methods and Results,” in Proc. IEEE/CVF CVPR Workshops, 2025, pp. 1525–1535. doi · arXiv
  61. Le Zhang, Ao Li, Qibin Hou, Ce Zhu, Yonina C. Eldar, “Deep-Learning-Empowered Super Resolution: A Comprehensive Survey and Future Prospects,” Proceedings of the IEEE, 2025. source
  62. W. Ren, R. Goyal, Z. Hu, T. T. Aumentado-Armstrong, I. Mohomed and A. Levinshtein, “Hallucination Score: Towards Mitigating Hallucinations in Generative Image Super-Resolution,” arXiv:2507.14367, 2025. arXiv
  63. Real-ESRGAN contributors, “Real-ESRGAN model and ×4-plus training configuration,” official source, accessed 2026-11-02. model · configuration
  64. D. P. Kingma, J. Ba, “Adam: A Method for Stochastic Optimization,” ICLR, 2015. source
  65. M. Bevilacqua, A. Roumy, C. Guillemot and M.-L. Alberi Morel, “Low-Complexity Single-Image Super-Resolution based on Nonnegative Neighbor Embedding,” in Proc. BMVC, 2012, pp. 135.1–135.10. doi
  66. R. Zeyde, M. Elad and M. Protter, “On Single Image Scale-Up Using Sparse-Representations,” in Curves and Surfaces, Lecture Notes in Computer Science, Springer, 2012, pp. 711–730. doi
  67. D. Martin, C. Fowlkes, D. Tal and J. Malik, “A Database of Human Segmented Natural Images and Its Application to Evaluating Segmentation Algorithms and Measuring Ecological Statistics,” in Proc. IEEE ICCV, 2001, vol. 2, pp. 416–423. doi
  68. J.-B. Huang, A. Singh and N. Ahuja, “Single Image Super-Resolution from Transformed Self-Exemplars,” in Proc. IEEE CVPR, 2015, pp. 5197–5206. doi · CVF
  69. J. Kim, J. K. Lee and K. M. Lee, “Accurate Image Super-Resolution Using Very Deep Convolutional Networks,” in Proc. IEEE CVPR, 2016, pp. 1646–1654. doi · arXiv
  70. C. Ledig, L. Theis, F. Huszár, J. Caballero, A. Cunningham, A. Acosta, A. Aitken, A. Tejani, J. Totz, Z. Wang and W. Shi, “Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network,” in Proc. IEEE CVPR, 2017, pp. 105–114. doi · arXiv
  71. R. Timofte, E. Agustsson, L. Van Gool, M.-H. Yang, L. Zhang et al., “NTIRE 2017 Challenge on Single Image Super-Resolution: Methods and Results,” in Proc. IEEE CVPR Workshops, 2017, pp. 1110–1121. doi · CVF
  72. Y. Matsui, K. Ito, Y. Aramaki, A. Fujimoto, T. Ogawa, T. Yamasaki and K. Aizawa, “Sketch-Based Manga Retrieval Using Manga109 Dataset,” Multimedia Tools and Applications, vol. 76, no. 20, pp. 21811–21838, 2017. doi
  73. KernelGAN contributors, “Adversarial losses and kernel constraints,” official source, accessed 2026-11-02. loss definitions · alternating updates
  74. X. Mao, Q. Li, H. Xie, R. Y. K. Lau, Z. Wang, S. P. Smolley, “Least Squares Generative Adversarial Networks,” ICCV, 2017. source
  75. DASR contributors, “MoCo logits and contrastive cross-entropy implementation,” official DASR source, commit 51524af8021c94a502ee002e135867a1b75dc7d0, accessed 2026-11-02. encoder · loss
  76. Y. Zhang, K. Li, K. Li, L. Wang, B. Zhong and Y. Fu, “Image Super-Resolution Using Very Deep Residual Channel Attention Networks,” in Proc. ECCV, 2018, pp. 294–310. doi · arXiv
  77. K. Simonyan and A. Zisserman, “Very Deep Convolutional Networks for Large-Scale Image Recognition,” in Proc. ICLR, 2015. arXiv
  78. Y. Zhang, X. Huang, J. Ma, et al., “Recognize Anything: A Strong Image Tagging Model,” CVPR Workshops, 2024. source
  79. E. J. Hu, Y. Shen, P. Wallis, Z. Allen-Zhu, Y. Li, S. Wang, L. Wang, W. Chen, “LoRA: Low-Rank Adaptation of Large Language Models,” ICLR, 2022. source
  80. D. Podell, Z. English, K. Lacey, A. Blattmann, T. Dockhorn, J. Müller, J. Penna, R. Rombach, “SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis,” ICLR, 2024. source
  81. H. Liu, C. Li, Q. Wu, Y. J. Lee, “Visual Instruction Tuning,” NeurIPS, 2023. source
  82. T. Karras, M. Aittala, T. Aila, S. Laine, “Elucidating the Design Space of Diffusion-Based Generative Models,” NeurIPS, 2022. source