FANO: Anchored Forward-Noise Optimization for Diffusion Inverse Problems
Organizations: College of Science, National University of Defense Technology
Abstract
Pre-trained diffusion models have become a widely adopted zero-shot prior for image inverse problems, among which annealed decoupling methods achieve leading reconstruction performance on nonlinear tasks. We find, however, that annealed methods that directly optimize the denoiser input exhibit radial drift in later layers: within-layer optimization drives the noise-component norm of away from its typical value, pushing the denoiser out of its high-performance regime and degrading reconstruction quality. To address this, we propose FANO (Anchored Forward-Noise Optimization). At each annealing layer, FANO fixes the previous denoiser estimate as an anchor and optimizes only the forward noise , whose distribution is known, instead of , thereby preserving the forward-diffusion structure of the denoiser input. The standard Gaussian prior on concentrates its norm near the typical value, and deviations degrade denoiser performance. We accordingly decompose in polar coordinates into a direction and a radial magnitude, and apply a soft potential to regulate the radius, keeping in the high-density region of the forward distribution. On multiple linear and nonlinear inverse problems, FANO matches the strongest current baselines on linear tasks and achieves consistent improvements on nonlinear tasks.
Figures & tables
| FFHQ | ImageNet | ||||||
| Task | Method | PSNR | SSIM | LPIPS | PSNR | SSIM | LPIPS |
| SR | DAPS | 29.35 1.47 | 0.782 0.028 | 0.268 0.035 | 25.47 2.86 | 0.638 0.087 | 0.376 0.061 |
| SITCOM | 30.55 1.87 | 0.868 0.033 | 0.223 0.034 | 27.09 3.84 | 0.757 0.116 | 0.307 0.077 | |
| FAST-DIPS | 29.59 1.75 | 0.843 0.041 | 0.241 0.042 | 26.55 3.65 | 0.725 0.125 | 0.325 0.072 | |
| DiffRGD | 29.24 1.93 | 0.838 0.042 | 0.214 0.037 | 25.96 3.70 | 0.712 0.136 | 0.322 0.084 | |
| INO | 29.24 1.80 | 0.809 0.037 | 0.260 0.038 | 25.65 3.50 | 0.694 0.131 | 0.310 0.090 | |
| Component | HDR | NL deblur | ||||||
|---|---|---|---|---|---|---|---|---|
| Variant | Polar | Radial freedom | PSNR | SSIM | LPIPS | PSNR | SSIM | LPIPS |
| FANO | ✓ | ✓ | 29.88 | 0.878 | 0.157 | 30.83 | 0.873 | 0.183 |
| FANO-Sphere | ✓ | 29.06 | 0.858 | 0.167 | 30.72 | 0.870 | 0.183 | |
| FANO-Euc | ✓ | 29.18 | 0.859 | 0.171 | 30.80 | 0.871 | 0.182 | |
Appendix figures & tables12 assets
Supplementary material from the paper’s appendix.
Appendix
| Task | Type | Parameters |
|---|---|---|
| SR | Linear | Bicubic, factor |
| Gauss. deblur | Linear | Gaussian, |
| Motion deblur | Linear | motion, intensity |
| Inpaint (box) | Linear | random mask |
| Inpaint (rand.) | Linear | Pixel mask prob. |
| Phase retrieval | Nonlinear | Oversampling ratio |
| Task | Annealing layers | Inner steps | |
|---|---|---|---|
| SR | 50 | 30 | |
| Gauss. deblur | 50 | 30 | |
| Motion deblur | 50 | 30 | |
| Inpaint (box) | 20 | 20 | |
| Inpaint (rand.) | 20 | 30 | |
| Phase retrieval | 50 | 30 |
| FFHQ | ImageNet | |||||
|---|---|---|---|---|---|---|
| Task | SITCOM | DAPS | Sphere | SITCOM | DAPS | Sphere |
| SR | 30.55 | 29.35 | 30.57 | 27.09 | 25.47 | 27.14 |
| Gauss. deblur | 29.89 | 29.61 | 30.66 | 26.38 | 25.88 | 27.01 |
| Motion deblur | 32.37 | 31.65 | 33.03 | 29.38 | 28.65 | 29.89 |
| Inpaint (box) | 24.47 | 24.32 | 25.14 | 19.62 | 21.05 | 20.17 |
| Inpaint (rand.) | 32.40 | 30.72 | 34.06 | 28.90 | 27.35 | 29.90 |
| FFHQ | ImageNet | ||||||
|---|---|---|---|---|---|---|---|
| Task | Method | PSNR | SSIM | LPIPS | PSNR | SSIM | LPIPS |
| SR | SITCOM-TB | 30.61 | 0.869 | 0.226 | 27.11 | 0.759 | 0.311 |
| FANO | 30.68 | 0.872 | 0.232 | 27.03 | 0.755 | 0.321 | |
| Gauss. deblur | SITCOM-TB | 30.09 | 0.854 | 0.244 | 26.56 | 0.732 | 0.337 |
| FANO | 30.65 | 0.868 | 0.225 | 27.01 | 0.746 | 0.317 | |
| Motion deblur | SITCOM-TB | 32.62 | 0.902 | 0.187 | 29.68 | 0.838 | 0.231 |
| Task | DAPS | FANO |
|---|---|---|
| SR | 83 | 74 |
| Gauss. deblur | 68 | 88 |
| Motion deblur | 68 | 88 |
| Inpaint (box) | 46 | 86 |
| Inpaint (rand.) | 46 | 34 |
| Phase retrieval | 107 | 87 |
| FFHQ | ImageNet | ||||||
|---|---|---|---|---|---|---|---|
| Task | Config | PSNR | SSIM | LPIPS | PSNR | SSIM | LPIPS |
| SR | 30.68 | 0.872 | 0.232 | 27.03 | 0.755 | 0.321 | |
| 30.64 | 0.869 | 0.226 | 27.10 | 0.756 | 0.315 | ||
| Gauss. deblur | 30.65 | 0.868 | 0.225 | 27.01 | 0.746 | 0.317 | |
| 30.55 | 0.866 | 0.217 | 27.08 | 0.750 | 0.306 | ||
| Motion deblur | 33.04 | 0.909 | 0.167 | 29.88 | 0.840 | 0.225 | |
Diffusion-Based Inverse Problems
Diffusion-based inverse problems use generative priors to recover images and physical fields from incomplete, noisy, or indirect measurements.
See all 156 papers