Real-world image super-resolution often seeks the most detailed plausible reconstruction, although more detail is not always desirable. The same degraded image may require restrained restoration, faithful reconstruction, or enhanced visual detail according to user intent. We present DetailScale-SR, a preference-conditioned framework that continuously controls reconstructed detail for a fixed low-resolution input without altering the scene being restored. Rather than relying on post-processing or repeatedly rewriting prompts, it learns a continuous family of reconstructions while separating content recovery from detail preference. This enables a smooth transition from conservative restoration to richly textured results under consistent input and semantic constraints. Its controllable range also extends beyond conventional restoration: DetailScale-SR can enhance high-frequency structures and generate perceptually richer textures than the paired high-resolution reference. Text restoration is another distinct strength. For text-rich scenes, especially Chinese posters and signs, it recovers more legible and character-faithful text than prior generative SR methods under matched textual guidance. DetailScale-SR therefore unifies continuous detail control, beyond-reference enhancement, and reliable recovery of semantically important content. Our model is available at https://modelscope.cn/models/DiffSynth-Studio/Qwen-Image-2.1-SR
Figures & tables
Figure 2: Overview of the proposed DetailScale-SR framework.
Dataset
Method
PSNR ↑
SSIM ↑
LPIPS ↓
DISTS ↓
FID ↓
MUSIQ ↑
MANIQA ↑
Rank ↓
RealSR
OSEDiff
25.15
0.7341
0.2921
0.2128
123.49
69.09
0.6326
7.70
PiSA-SR
25.50
0.7417
0.2672
0.2044
124.09
70.15
0.6560
5.80
TSD-SR
24.81
0.7172
0.2743
0.2104
114.45
71.19
0.6347
8.00
TVT
25.81
0.7596
0.2597
0.2061
109.90
69.89
0.6232
4.40
SUPIR
23.65
0.6620
0.3541
0.2488
130.38
62.09
0.5780
11.80
DiT4SR
23.50
0.6657
0.3215
0.2251
118.55
67.76
0.6564
10.60
Table 1: Quantitative comparison on RealSR (top) and DRealSR (bottom). Best and second-best values of each column are highlighted in bold and underlined. Rank is the average per-metric rank over PSNR, SSIM, LPIPS, DISTS and FID (1 = best, lower is better); the no-reference metrics are reported but not ranked, as they measure perceived quality rather than agreement with the reference.
Figure 3: Qualitative comparison on ×4 real-world super-resolution.Our method performs favorably overall and is especially strong on text-rich content.
Figure 4: Controllable SR results on two heavily degraded images; Our method delivers clearer detail-controllable generation and preserves structure better under heavy degradation.
Setting
PSNR ↑
SSIM ↑
LPIPS ↓
MANIQA ↑
MUSIQ ↑
FID ↓
s=0.0
26.69
0.7723
0.2766
0.5533
60.94
130.46
s=0.4
26.56
0.7685
0.2292
0.6030
63.00
98.30
s=0.8
26.14
0.7514
0.2215
0.6251
64.26
87.44
s=1.0
25.48
0.7300
0.2343
0.6368
65.53
88.40
CFG γ=2
24.06
0.6587
0.2835
0.6653
67.86
98.53
s=0.0
29.08
0.8243
0.2985
0.5346
57.01
147.76
Table 2: Continuous detail control on RealSR (top) and DRealSR (bottom). The last row of each block applies Detail-CFG.
Figure 5: Text restoration with prompt assistance. Our method performs favorably on challenging cases containing many texts and recovers the content accurately; more examples are provided in the supplementary material.
Strategy
PSNR ↑
SSIM ↑
LPIPS ↓
DISTS ↓
FID ↓
Two-stage (frozen)
25.48
0.7300
0.2343
0.1834
88.40
Joint (one-stage)
23.50
0.6605
0.2972
0.2171
102.77
Table 3: Ablation on the two-stage training strategy on RealSR
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.
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
Real-world image super-resolution (Real-ISR) requires balancing structural fidelity to degraded observations with realistic detail synthesis. However, existing generative Real-ISR methods often rely on entangled conditioning mechanisms, leading to structural drift or semantically inconsistent details. To address this issue, we propose Visual In-Context Restoration (VICR), a Diffusion Transformer (DiT)-based framework that formulates Real-ISR as image completion. Specifically, we introduce a decoupled visual prior injection mechanism that derives local and global cues from the low-quality (LQ) image: local cues help recover image structures and support high-frequency detail synthesis, while global cues guide overall generation and promote semantic consistency. For ambiguous regions under severe degradation, VICR employs an inference-time agent to refine semantic prompts using visual evidence from the LQ input while keeping model parameters fixed. Experiments show that VICR achieves state-of-the-art performance across multiple Real-ISR benchmarks with only 127M trainable parameters.
Qichang Zhang, Hailong Wang, Baiang Li +4
Faculty of Science and Technology, University of Macau · 2Nullmax · 3Hefei University of Technology +1