Why a posterior sample costs twice the MMSE
Optional · supplementary note 4 of 6 for DLCV 09A · Non-Blind Image Super-Resolution. Nothing in the main text depends on it; 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
This note derives the factor of 2 between the MSE of a posterior sample and that of the posterior mean.
Plain version. The posterior mean sits exactly in the middle of all plausible answers. The true image is one random point in that cloud, and a posterior sample is another, independent random point in the same cloud. Two random points are, on average, twice as far apart (in squared distance) as one random point is from the center.
Precise version. Let be the HR image, the LR observation, the posterior mean, and a sample drawn from independently of given . Then, conditioned on ,
where is the Euclidean norm over all pixels. The second line expands the square. In the third, the two squared terms are equal because and have the same conditional distribution, and the cross term vanishes because, given , the two are independent and each has mean . Taking the expectation over gives : the sampler’s MSE is exactly twice the minimum MSE, about a 3 dB gap when PSNR is computed from the expected MSE. Averaging per-image PSNR is a different statistic and need not give that exact gap. Its output distribution equals the distribution of real images, so it achieves perfect perceptual quality under this distributional definition; this does not guarantee that a sampled detail belongs to the particular photographed scene. Blau and Michaeli use this construction to show that perfect perceptual quality never costs more than a factor of 2 in MSE [2].
The same algebra gives the curve in Figure 1. For the blend with ,
where the cross term vanishes for the same reason. The distortion grows smoothly from the MMSE value at to twice that value at , while the blend’s outputs move from the blurry posterior mean toward realistic samples. In our toy, the measured MSE went from 0.0072 to 0.0143.
Code · Python · 8 lines
import numpy as np
rng = np.random.default_rng(1)
mu = rng.standard_normal((100000, 1)) # posterior means for many observations y
X = mu + rng.standard_normal(mu.shape) # the truth: one draw from p(x | y) = N(mu, 1)
X_hat = mu + rng.standard_normal(mu.shape) # an independent posterior sample
for a in (0.0, 0.5, 1.0):
print(f"alpha {a:.1f}: MSE {np.mean((X - (mu + a * (X_hat - mu))) ** 2):.3f} (theory {1 + a * a:.3f})")Output
alpha 0.0: MSE 1.002 (theory 1.000)
alpha 0.5: MSE 1.250 (theory 1.250)
alpha 1.0: MSE 1.996 (theory 2.000)References
- S. C. Park, M. K. Park and M. G. Kang, “Super-Resolution Image Reconstruction: A Technical Overview,” IEEE Signal Processing Magazine, vol. 20, no. 3, pp. 21–36, 2003. doi
- Y. Blau and T. Michaeli, “The Perception-Distortion Tradeoff,” in Proc. IEEE/CVF CVPR, 2018, pp. 6228–6237. doi · arXiv
- MathWorks, “imresize: Resize image,” MATLAB documentation (default method bicubic; antialiasing on when shrinking). link
- Z. Wang, J. Chen and S. C. H. Hoi, “Deep Learning for Image Super-Resolution: A Survey,” IEEE TPAMI, vol. 43, no. 10, pp. 3365–3387, 2021. doi · arXiv
- R. Keys, “Cubic Convolution Interpolation for Digital Image Processing,” IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 29, no. 6, pp. 1153–1160, 1981. doi
- C. E. Duchon, “Lanczos Filtering in One and Two Dimensions,” Journal of Applied Meteorology, vol. 18, no. 8, pp. 1016–1022, 1979. doi
- W. T. Freeman, T. R. Jones and E. C. Pasztor, “Example-Based Super-Resolution,” IEEE Computer Graphics and Applications, vol. 22, no. 2, pp. 56–65, 2002. doi
- D. Glasner, S. Bagon and M. Irani, “Super-Resolution from a Single Image,” in Proc. IEEE ICCV, 2009, pp. 349–356. doi
- J. Yang, J. Wright, T. S. Huang and Y. Ma, “Image Super-Resolution via Sparse Representation,” IEEE Transactions on Image Processing, vol. 19, no. 11, pp. 2861–2873, 2010. doi
- R. Timofte, V. De Smet and L. Van Gool, “Anchored Neighborhood Regression for Fast Example-Based Super-Resolution,” in Proc. IEEE ICCV, 2013, pp. 1920–1927. doi · CVF
- R. Timofte, V. De Smet and L. Van Gool, “A+: Adjusted Anchored Neighborhood Regression for Fast Super-Resolution,” in Computer Vision – ACCV 2014, Lecture Notes in Computer Science, Springer, 2015, pp. 111–126. doi
- Y. Romano, J. Isidoro and P. Milanfar, “RAISR: Rapid and Accurate Image Super Resolution,” IEEE Transactions on Computational Imaging, vol. 3, no. 1, pp. 110–125, 2017. doi · arXiv
- 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
- C. Dong, C. C. Loy, K. He and X. Tang, “Image Super-Resolution Using Deep Convolutional Networks,” IEEE TPAMI, vol. 38, no. 2, pp. 295–307, 2016. doi · arXiv
- C. Dong, C. C. Loy and X. Tang, “Accelerating the Super-Resolution Convolutional Neural Network,” in Proc. ECCV, 2016, pp. 391–407. doi · arXiv
- W. Shi, J. Caballero, F. Huszár, J. Totz, A. P. Aitken, R. Bishop, D. Rueckert and Z. Wang, “Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network,” in Proc. IEEE CVPR, 2016, pp. 1874–1883. doi · arXiv
- 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
- B. Lim, S. Son, H. Kim, S. Nah and K. M. Lee, “Enhanced Deep Residual Networks for Single Image Super-Resolution,” in Proc. IEEE CVPR Workshops, 2017, pp. 1132–1140. doi · arXiv
- 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
- J. Liang, J. Cao, G. Sun, K. Zhang, L. Van Gool and R. Timofte, “SwinIR: Image Restoration Using Swin Transformer,” in Proc. IEEE/CVF ICCV Workshops, 2021, pp. 1833–1844. doi · arXiv
- X. Chen, X. Wang, J. Zhou, Y. Qiao and C. Dong, “Activating More Pixels in Image Super-Resolution Transformer,” in Proc. IEEE/CVF CVPR, 2023, pp. 22367–22377. doi · arXiv
- PyTorch contributors, “PixelShuffle.cpp” (implementation of
torch.nn.PixelShuffleandtorch.nn.PixelUnshuffle), PyTorch source code. GitHub - A. Aitken, C. Ledig, L. Theis, J. Caballero, Z. Wang and W. Shi, “Checkerboard Artifact Free Sub-Pixel Convolution: A Note on Sub-Pixel Convolution, Resize Convolution and Convolution Resize,” arXiv:1707.02937, 2017. arXiv
- 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
- 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
- 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
- Z. Liu, Y. Lin, Y. Cao, H. Hu, Y. Wei, Z. Zhang, S. Lin and B. Guo, “Swin Transformer: Hierarchical Vision Transformer using Shifted Windows,” in Proc. IEEE/CVF ICCV, 2021, pp. 9992–10002. doi · arXiv
- Z. Chen, Y. Zhang, J. Gu, L. Kong, X. Yang and F. Yu, “Dual Aggregation Transformer for Image Super-Resolution,” in Proc. IEEE/CVF ICCV, 2023, pp. 12278–12287. doi · arXiv
- H. Guo, J. Li, T. Dai, Z. Ouyang, X. Ren and S.-T. Xia, “MambaIR: A Simple Baseline for Image Restoration with State-Space Model,” in Proc. ECCV, 2024, pp. 222–241. doi · arXiv
- K. Simonyan and A. Zisserman, “Very Deep Convolutional Networks for Large-Scale Image Recognition,” in Proc. ICLR, 2015. arXiv
- J. Johnson, A. Alahi and L. Fei-Fei, “Perceptual Losses for Real-Time Style Transfer and Super-Resolution,” in Proc. ECCV, 2016, pp. 694–711. doi · arXiv
- Y. Blau, R. Mechrez, R. Timofte, T. Michaeli and L. Zelnik-Manor, “The 2018 PIRM Challenge on Perceptual Image Super-Resolution,” in Proc. ECCV Workshops, 2018. arXiv
- 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
- C. Ma, C.-Y. Yang, X. Yang and M.-H. Yang, “Learning a No-Reference Quality Metric for Single-Image Super-Resolution,” Computer Vision and Image Understanding, vol. 158, pp. 1–16, 2017. doi · arXiv
- C. Saharia, J. Ho, W. Chan, T. Salimans, D. J. Fleet and M. Norouzi, “Image Super-Resolution via Iterative Refinement,” IEEE TPAMI, 2022. doi · arXiv
- O. Ronneberger, P. Fischer and T. Brox, “U-Net: Convolutional Networks for Biomedical Image Segmentation,” in Proc. MICCAI, 2015, pp. 234–241. doi · arXiv
- X. Hu, H. Mu, X. Zhang, Z. Wang, T. Tan and J. Sun, “Meta-SR: A Magnification-Arbitrary Network for Super-Resolution,” in Proc. IEEE/CVF CVPR, 2019, pp. 1575–1584. doi · arXiv
- Y. Chen, S. Liu and X. Wang, “Learning Continuous Image Representation with Local Implicit Image Function,” in Proc. IEEE/CVF CVPR, 2021, pp. 8624–8634. doi · arXiv
- Z. Hui, X. Gao, Y. Yang and X. Wang, “Lightweight Image Super-Resolution with Information Multi-distillation Network,” in Proc. ACM International Conference on Multimedia, 2019, pp. 2024–2032. doi · arXiv
- B. Ren, Y. Li, N. Mehta, R. Timofte et al., “The Ninth NTIRE 2024 Efficient Super-Resolution Challenge Report,” in Proc. IEEE/CVF CVPR Workshops, 2024. CVF · arXiv
- B. Ren, H. Guo, L. Sun, Z. Wu, R. Timofte, Y. Li et al., “The Tenth NTIRE 2025 Efficient Super-Resolution Challenge Report,” in Proc. IEEE/CVF CVPR Workshops, 2025. CVF · arXiv
- ITU-R, “Recommendation BT.601: Studio Encoding Parameters of Digital Television for Standard 4:3 and Wide Screen 16:9 Aspect Ratios.” ITU
- MathWorks, “rgb2ycbcr: Convert RGB color values to YCbCr color space,” Image Processing Toolbox documentation. link
- 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
- Z. Chen, Z. Wu, E. Zamfir, K. Zhang, Y. Zhang, R. Timofte et al., “NTIRE 2024 Challenge on Image Super-Resolution (×4): Methods and Results,” in Proc. IEEE/CVF CVPR Workshops, 2024, pp. 6108–6132. doi · arXiv
- 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
- 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
- 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
- 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
- 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
- A. Fujimoto, T. Ogawa, K. Yamamoto, Y. Matsui, T. Yamasaki and K. Aizawa, “Manga109 Dataset and Creation of Metadata,” in Proc. 1st International Workshop on coMics ANalysis, Processing and Understanding (MANPU), 2016, pp. 1–5. doi
- 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
- 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
- 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
- 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
- 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
- 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