Real-world image super-resolution (SR) requires recovering perceptually realistic high-resolution images from complex low-resolution observations while preserving faithful content. Diffusion-based SR benefits from strong generative priors but incurs substantial computational overhead, whereas feed-forward CNN and Transformer SR models are efficient yet often struggle to recover realistic high-frequency details. This motivates a natural question: can diffusion priors be transferred to existing diffusion-free SR networks without introducing diffusion components at inference time? To this end, we propose PhoenixSR, a generative heterogeneous distillation framework that transfers diffusion priors to independently designed feed-forward SR networks through score-based distribution matching. Rather than aligning heterogeneous features or imitating sampled diffusion outputs, PhoenixSR uses the pretrained diffusion model as distribution-level supervision, while paired SR supervision preserves reconstruction fidelity. To make distribution matching effective for fidelity-sensitive SR, we introduce Heterogeneous Distribution Adaptation, which adapts the target score to the SR domain, improves tracking of the evolving student distribution, and anchors training with paired supervision. We further employ Directional Reliability Weighting, a lightweight residual-consistency-based reweighting strategy that reduces unstable distributional guidance. All diffusion-related components are removed after training, leaving the original student architecture and inference cost unchanged. Experiments on three SR benchmarks and six feed-forward backbones, including SwinIR, HAT, Real-ESRGAN, and SeeMoRe, show consistent perceptual improvements with largely preserved reconstruction fidelity.
Figures & tables
Figure 1: Overview and performance of PhoenixSR. (a) PhoenixSR transfers diffusion generative priors to diffusion-free SR students through generative heterogeneous distillation. (b) Visual comparisons of SwinIR and HAT before and after PhoenixSR distillation. (c) Average and model-wise performance improvements brought by PhoenixSR on RealSR.
Figure 2: Motivation. CKA ( Kornblith et al., 2019 ) (left) reveals weak feature correspondence between DBSR and DFSR, while residual and zoom-in comparisons (right) show local discrepancies between DBSR outputs and paired HR targets.
Figure 3: Effect of fidelity anchoring on distribution matching. (a) Vanilla DMD causes unstable optimization and rapid PSNR degradation when directly applied to SR. (b) Sample-level fidelity supervision stabilizes distribution matching and preserves reconstruction quality. The optimization landscapes are conceptual illustrations.
Figure 4: Overview of PhoenixSR. HDA adapts real-score estimation, tracks the evolving student distribution, and anchors paired fidelity. DRW reweights sample-wise guidance according to residual-direction consistency.
Figure 5: t-SNE projection of feature representations. After distillation, DFSR representations move toward the region between real HR samples and DBSR outputs in t-SNE projection.
Model
PSNR ↑
LPIPS ↓
MANIQA ↑
NIQE ↓
MUSIQ ↑
CLIP-IQA ↑
LIQE-Mix ↑
ARNIQA ↑
TOPIQ ↑
Params
FLOPs
Runtime
(M)
(G)
(ms)
DBSR Models
StableSR
24.4272
0.2636
0.5910
6.5779
61.8041
0.5644
3.5794
0.6772
0.5121
1410
162731
17870
SeeSR
25.1475
0.3007
0.6451
5.3958
69.8181
0.6703
3.6576
0.7152
0.6888
2511
131980
5451
SinSR
26.1309
0.3068
0.5432
6.1038
61.2905
0.6236
3.5595
0.6601
0.5309
119
5392
201
OSEDiff
25.1554
0.2928
0.6343
5.6243
69.1793
0.6669
4.1382
0.6935
0.6292
1761
4611
169
Table 1: Quantitative comparison on RealSR. PhoenixSR is applied to DFSR backbones, while DBSR methods are included as reference baselines. Gray cells denote results with PhoenixSR; red and blue denote performance improvement and degradation, respectively.
Model
PSNR ↑
LPIPS ↓
MANIQA ↑
NIQE ↓
MUSIQ ↑
CLIP-IQA ↑
LIQE-Mix ↑
ARNIQA ↑
TOPIQ ↑
DBSR Models
StableSR
28.4070
0.2726
0.4914
7.5141
51.4058
0.5046
2.6971
0.5762
0.4321
SeeSR
28.0712
0.3174
0.6052
6.4061
65.0873
0.6909
3.6846
0.6880
0.6572
SinSR
28.2012
0.3511
0.5034
6.7220
57.5658
0.6574
3.5766
0.6224
0.5288
OSEDiff
27.9321
0.2971
0.5874
6.5227
64.8184
0.6909
3.9768
0.6702
0.6005
AdcSR
26.2041
0.3156
0.5937
6.1738
66.8444
0.7206
4.2538
0.6985
0.6602
Table 2: Quantitative comparison on DRealSR.
Model
PSNR ↑
LPIPS ↓
MANIQA ↑
NIQE ↓
MUSIQ ↑
CLIP-IQA ↑
LIQE-Mix ↑
ARNIQA ↑
TOPIQ ↑
DBSR Models
StableSR
24.2395
0.3129
0.5506
5.6054
58.4136
0.5627
3.4477
0.6580
0.4851
SeeSR
23.6780
0.3194
0.6222
4.8095
68.6710
0.6935
4.0053
0.7301
0.6853
SinSR
24.2883
0.3222
0.5414
5.8005
63.2651
0.6530
3.7998
0.6745
0.5796
OSEDiff
23.7214
0.2942
0.6130
4.7203
67.9926
0.6611
4.2976
0.7030
0.6194
AdcSR
23.3567
0.2845
0.6071
4.3420
68.0922
0.6846
4.3009
0.7215
0.6574
Table 3: Quantitative comparison on DIV2K-Val.
Figure 6: Qualitative comparison on DIV2K-Val. PhoenixSR enhances high-frequency details of diffusion-free SR models, producing clearer and more natural wood textures while preserving the underlying image structure.
Figure 7: Qualitative comparison with diffusion-based SR methods. Diffusion-based baselines may over-sharpen the statue surface or introduce hallucinated textures that deviate from the underlying geometry. PhoenixSR improves details while retaining more faithful structural reconstruction.
SwinIR-L
SwinIR-S
SeeMoRe
Method
PSNR ↑
LPIPS ↓
MUSIQ ↑
PSNR ↑
LPIPS ↓
MUSIQ ↑
PSNR ↑
LPIPS ↓
MUSIQ ↑
Task-only strong FT
26.41
0.257
59.7
25.13
0.301
56.0
26.00
0.289
53.7
Output KD + GT
25.97
0.274
56.7
24.77
0.327
55.8
25.98
0.301
52.1
SDS + SFA
25.77
0.273
56.5
24.89
0.305
50.4
25.10
0.317
51.8
DMD2 + SFA
25.55
0.256
57.8
24.78
0.311
54.7
24.87
0.304
53.6
PhoenixSR w/o DRW
26.15
0.247
63.5
24.79
0.296
61.5
25.35
0.290
60.8
Table 4: Comparison of alternative diffusion-to-DFSR transfer strategies on RealSR.
Variant
SFA
Score Adapt.
DRW
PSNR ↑
MANIQA ↑
MUSIQ ↑
CLIP-IQA ↑
Base SwinIR-L
–
–
–
26.3071
0.5228
58.6388
0.4502
SFA only
✓
–
–
26.3124
0.5067
56.8245
0.4538
w/o SFA
–
✓
✓
20.7107
0.3512
40.7941
0.3071
SFA + Score Adapt.
✓
✓
–
26.1587
0.5678
63.5729
0.4963
Full PhoenixSR
✓
✓
✓
26.4409
0.5745
64.3102
0.5026
Table 5: Ablation of the main PhoenixSR components.
Appendix figures & tables10 assets
Supplementary material from the paper’s appendix.
Appendix
Method
Task
Transfer mechanism
Deployed model
Native DFSR
Diffusion-derived inference
DTKD ( Park & Song, 2026 )
SR
Frequency-aware output KD
Transformer SR student
✓
No
DM-SR ( Park & Lee, 2026 )
SR
Distribution alignment
Pretrained diffusion model with learned input mapping
–
Yes
Cross-Space Distillation ( Nguyen et al., 2026 )
Image generation
Distribution-level diffusion distillation
One-step diffusion student
–
Yes
PhoenixSR
SR
Score-based distribution matching
CNN / Transformer / lightweight SR students
✓
No
Appendix
Table 6: Positioning of PhoenixSR with closely related diffusion-transfer methods. “Native DFSR” denotes a conventional feed-forward SR architecture designed independently from the diffusion teacher. “Diffusion-derived inference” indicates whether the deployed model still relies on a diffusion model or a diffusion-derived generator.
Figure 8: Analysis of diffusion guidance for SR distillation. (a) Effect of timestep selection. Low-noise states preserve more input-consistent structures, while stronger noise makes diffusion guidance increasingly dominated by the generative prior and may introduce unfaithful details. (b) Visualization of real-score adaptation. The frozen real-score model may convert restoration artifacts into plausible but unfaithful textures, while LoRA adaptation alleviates such artifact-to-texture degeneration and provides more SR-compatible guidance.
Student
Method
PSNR ↑
LPIPS ↓
MANIQA ↑
MUSIQ ↑
CLIP-IQA ↑
SwinIR-L
Base checkpoint
26.30
0.253
0.522
58.6
0.45
PhoenixSR w/o DRW
26.20 ± 0.03
0.251 ± 0.004
0.557 ± 0.007
62.9 ± 0.3
0.481 ± 0.005
PhoenixSR
26.40 ± 0.04
0.247 ± 0.008
0.568 ± 0.006
64.1 ± 0.5
0.493 ± 0.007
SeeMoRe
Base checkpoint
26.0
0.295
0.347
48.7
0.381
PhoenixSR w/o DRW
25.41 ± 0.07
0.277 ± 0.009
0.511 ± 0.008
60.9 ± 0.4
0.519 ± 0.009
PhoenixSR
25.94 ± 0.03
0.269 ± 0.005
0.531 ± 0.005
61.7 ± 0.3
0.530 ± 0.006
Appendix
Table 7: Robustness across random seeds on RealSR. PhoenixSR variants are independently trained with three random seeds and reported as mean ± standard deviation. The base checkpoints are fixed references evaluated once.
Table 16Table 17Table 18
Figure 9: Additional qualitative comparisons on RealSR. PhoenixSR improves fine structures and local texture reconstruction across different feed-forward SR backbones while preserving the underlying scene content and geometry. Best viewed by zooming in.
Figure 10: Additional qualitative comparisons on DRealSR. PhoenixSR recovers clearer local structures and more natural texture details across different feed-forward SR backbones while largely preserving reconstruction fidelity. Best viewed by zooming in.
Figure 11: User study across six DFSR backbones.
Backbone
Original
PhoenixSR
Δ (95% CI)
Real-ESRGAN
6.38 [6.14, 6.64]
7.89 [7.71, 8.06]
+1.51 [1.36, 1.66]
HAT
6.78 [6.56, 7.00]
8.29 [8.15, 8.40]
+1.51 [1.39, 1.63]
SwinIR-L
6.67 [6.43, 6.92]
8.24 [8.10, 8.37]
+1.57 [1.44, 1.70]
SwinIR-M
6.22 [5.92, 6.51]
7.50 [7.31, 7.67]
+1.28 [1.14, 1.42]
SwinIR-S
6.08 [5.95, 6.21]
6.87 [6.60, 7.11]
+0.79 [0.63, 0.92]
SeeMoRe
5.74 [5.58, 5.89]
6.81 [6.60, 7.04]
+1.07 [0.87, 1.31]
Appendix
Table 14: Detailed user-study statistics. Scores are reported as image-level mean scores with 95% bootstrap confidence intervals. Δ denotes PhoenixSR minus the original backbone.
Diffusion-based approaches have recently driven remarkable progress in real-world image super-resolution (SR). However, existing methods still struggle to simultaneously preserve fine details and ensure high-fidelity reconstruction, often resulting in suboptimal visual quality. In this paper, we propose FiDeSR, a high-fidelity and detail-preserving one-step diffusion super-resolution framework. During training, we introduce a detail-aware weighting strategy that adaptively emphasizes regions where the model exhibits higher prediction errors. During inference, low- and high-frequency adaptive enhancers further refine the reconstruction without requiring model retraining, enabling flexible enhancement control. To further improve the reconstruction accuracy, FiDeSR incorporates a residual-in-residual noise refinement, which corrects prediction errors in the diffusion noise and enhances fine detail recovery. FiDeSR achieves superior real-world SR performance compared to existing diffusion-based methods, producing outputs with both high perceptual quality and faithful content restoration. The source code will be released at: https://github.com/Ar0Kim/FiDeSR.
Aro Kim, Myeongjin Jang, Chaewon Moon +3
Kyungpook National University · Korea Electronics Technology Institute
Large-scale pre-trained diffusion models have been extensively adopted for real-world image Super-Resolution because of their powerful generative priors through textual guidance. However, when super-resolving high-resolution images with patch-wise inference strategy, most existing diffusion-based SR methods tend to suffer from over-generation, due to the misalignment between the global prompt from LR image and the incomplete semantic information of local patches during each inference step. On the other hand, most existing methods also failed to generate detailed texture in local patches due to the overemphasis on global generation capabilities in network designs and training strategies. To address this issue, we present DreamSR, a novel SR model that suppresses local over-generation and improves fine-detail synthesis, thereby achieving visually faithful results with ultra-high-quality details. Specifically, we propose a dual-branch MM-ControlNet, where the ControlNet generates local textual feature with patch-level prompts while the pre-trained DiT provides global textual feature with global prompts, thereby mitigating over-generation and ensuring semantic consistency across patches. We also design a comprehensive training strategy with stage-specific data processing pipelines and a Receptive-Field Enhancement strategy, enhancing the model's capability to capture patch information and effectively restore local textures. Extensive experiments demonstrate that DreamSR outperforms state-of-the-art methods, providing high-quality SR results. Code and model are available at https://github.com/jerrydong0219/DreamSR.
Latent diffusion models have recently achieved remarkable progress in Real-world Image Super-Resolution (Real-SR). However, existing methods consistently exhibit a persistent seesaw effect: enhancing perceptual realism compromises structural fidelity. This effect, we argue, is not intrinsic to diffusion models, but largely stems from the architectural constraints of prevailing two-stage frameworks. To address this challenge, we propose PixelSR, a novel single-stage Pixel Diffusion Transformer for Real-SR that operates directly in pixel space. Specifically, PixelSR consists of two core designs: Pixel-space Conditional Diffusion and Orthogonal Flow Decoupling. The former establishes an effective pixel-space restoration framework by integrating spatially aligned LR conditioning, visual semantic guidance, and lightweight texture refinement, thereby bypassing VAE compression and enabling end-to-end optimization directly in RGB space. Complementarily, the latter decomposes the asymmetric flow target into a low-rank semantic component and a noise-free orthogonal texture component. By routing these components to specialized branches, it achieves decoupled optimization with reduced mutual interference. Benefiting from these designs, our PixelSR restores perceptually realistic details while faithfully preserving structural content. Extensive experiments across multiple benchmarks demonstrate that PixelSR not only achieves competitive performance, but also effectively mitigates the fidelity--realism seesaw effect.
Xiang Lv, Kai Zhao, Qiang Wang +3
China University of Petroleum (East China) · Kuaishou Technology · Shenzhen University of Advanced Technology