面向扩散图像超分辨率的解耦文本先验
Disentangled Textual Priors for Diffusion-based Image Super-Resolution
把文本先验按空间层级（全局/局部）与频率语义（低频/高频）解耦，并通过专用交叉注意力模块分阶段注入扩散采样过程，这一「解耦先验 + 频率感知负提示」的注入范式可直接迁移到其他条件扩散复原任务（去噪、去模糊、修复）以提升语义可控性。
原文依据 The corresponding embeddings are injected via specialized cross-attention modules, forming a progressive generation pipeline
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论图像超分辨率（SR）旨在从退化的低分辨率输入中重建高分辨率图像。尽管基于扩散的 SR 方法具备强大的生成能力，其性能在很大程度上取决于语义先验如何被组织并整合进生成过程。现有方法往往依赖纠缠或粗粒度的先验，将全局布局与局部细节混在一起，或把结构与纹理线索混为一谈，从而限制了语义可控性与可解释性。在本工作中，我们提出 DTPSR，一种新的基于扩散的 SR 框架，沿两个互补维度引入解耦文本先验：空间层级（全局与局部）与频率语义（低频与高频）。通过显式分离这些先验，DTPSR 使模型能够在频率感知的语义引导下同时捕捉场景级结构与物体特定细节。相应的嵌入通过专用交叉注意力模块注入，形成一条渐进式生成流水线，反映视觉内容从全局布局到细粒度纹理的语义粒度。为支撑这一范式，我们构建了 DisText-SR，一个大规模数据集，包含约 95,000 个图像-文本对，并配有精心解耦的全局、低频与高频描述。为进一步增强可控性与一致性，我们采用多分支无分类器引导策略，配合频率感知的负提示以抑制幻觉与语义漂移。在合成与真实世界基准上的大量实验表明，DTPSR 取得了较高的感知质量、有竞争力的保真度，并在多样退化场景下表现出强泛化能力。
英文原文摘要 1756 字符
Image Super-Resolution (SR) aims to reconstruct high-resolution images from degraded low-resolution inputs. While diffusion-based SR methods offer powerful generative capabilities, their performance heavily depends on how semantic priors are structured and integrated into the generation process. Existing approaches often rely on entangled or coarse-grained priors that mix global layout with local details, or conflate structural and textural cues, thereby limiting semantic controllability and interpretability. In this work, we propose DTPSR, a novel diffusion-based SR framework that introduces disentangled textual priors along two complementary dimensions: spatial hierarchy (global vs. local) and frequency semantics (low- vs. high-frequency). By explicitly separating these priors, DTPSR enables the model to simultaneously capture scene-level structure and object-specific details with frequency-aware semantic guidance. The corresponding embeddings are injected via specialized cross-attention modules, forming a progressive generation pipeline that reflects the semantic granularity of visual content--from global layout to fine-grained textures. To support this paradigm, we construct DisText-SR, a large-scale dataset containing approximately 95,000 image-text pairs with carefully disentangled global, low-frequency, and high-frequency descriptions. To further enhance controllability and consistency, we adopt a multi-branch classifier-free guidance strategy with frequency-aware negative prompts to suppress hallucinations and semantic drift. Extensive experiments on synthetic and real-world benchmarks show that DTPSR achieves high perceptual quality, competitive fidelity, and strong generalization across diverse degradation scenarios.