用频域对齐蒸馏，让扩散模型更清楚地还原图像细节。
FRAMER: Frequency-Aligned Self-Distillation with Adaptive Modulation Leveraging Diffusion Priors for Real-World Image Super-Resolution
- 通过频域分解，让每层特征学得更符合高频细节重建规律。
- 在多个模型上提升清晰度，尤其改善了边缘和纹理的还原效果。
- 适合做真实图像超分、想用扩散模型又怕细节丢失的研究者。
真实图像超分辨率（Real-ISR）旨在从带有混合未知退化的低分辨率输入中恢复高分辨率图像。尽管扩散模型在感知质量上超越生成对抗网络（GANs），但因存在低频偏倚及深度上的“先低频后高频”结构，导致高频（HF）细节重建不足。本文提出FRAMER，一种无需改动主干网络或推理流程的即插即用训练方案。在每个去噪步骤中，最终层特征图指导所有中间层学习。通过傅里叶变换掩码将教师与学生特征图分解为低频（LF）与高频（HF）成分，实现监督信号与模型内部频率层次的对齐。针对低频，采用类内对比损失（IntraCL）稳定全局共享结构；针对高频，引入类间对比损失（InterCL），利用随机层与批次内负样本强化实例特定细节。设计了两个自适应调制模块：基于频率的自适应权重（FAW）与频率对齐调制（FAM），分别重加权各层的高低频信号，并根据当前相似性动态控制蒸馏强度。在U-Net与DiT主干（如Stable Diffusion 2、3）上，FRAMER一致提升PSNR/SSIM及感知指标（LPIPS、NIQE、MANIQA、MUSIQ）。消融实验验证了最终层教师与随机层负样本的有效性。
原文摘要 · Abstract (English)
Real-image super-resolution (Real-ISR) seeks to recover HR images from LR inputs with mixed, unknown degradations. While diffusion models surpass GANs in perceptual quality, they under-reconstruct high-frequency (HF) details due to a low-frequency (LF) bias and a depth-wise "low-first, high-later" hierarchy. We introduce FRAMER, a plug-and-play training scheme that exploits diffusion priors without changing the backbone or inference. At each denoising step, the final-layer feature map teaches all intermediate layers. Teacher and student feature maps are decomposed into LF/HF bands via FFT masks to align supervision with the model's internal frequency hierarchy. For LF, an Intra Contrastive Loss (IntraCL) stabilizes globally shared structure. For HF, an Inter Contrastive Loss (InterCL) sharpens instance-specific details using random-layer and in-batch negatives. Two adaptive modulators, Frequency-based Adaptive Weight (FAW) and Frequency-based Alignment Modulation (FAM), reweight per-layer LF/HF signals and gate distillation by current similarity. Across U-Net and DiT backbones (e.g., Stable Diffusion 2, 3), FRAMER consistently improves PSNR/SSIM and perceptual metrics (LPIPS, NIQE, MANIQA, MUSIQ). Ablations validate the final-layer teacher and random-layer negatives.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。