Abstract
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.
Abstract (translated)
图像超分辨率（SR）的目标是从退化的低分辨率输入中重建高质量的高分辨率图像。虽然基于扩散的方法提供了强大的生成能力，但其性能很大程度上取决于语义先验如何被构建和整合到生成过程中。现有方法常常依赖于纠缠不清或粗粒度的先验信息，这些信息混合了全局布局与局部细节，或者混淆了结构和纹理线索，从而限制了语义控制能力和解释性。 在这项工作中，我们提出了DTPSR（Disentangled Textual Priors for Image Super-Resolution），这是一种新型基于扩散的方法框架。它在两个互补维度上引入了非纠缠的文本先验：空间层级（全局 vs. 局部）和频率语义（低频 vs. 高频）。通过明确分离这些先验，DTPSR使模型能够同时捕捉场景级别的结构和对象特定的细节，并且具有频率感知的语义引导。相应的嵌入信息通过专门设计的交叉注意力模块注入，形成了一个渐进生成管道，该管道从全局布局到精细纹理反映视觉内容的语义粒度。 为了支持这一范式，我们构建了DisText-SR，这是一个大规模的数据集，包含大约95,000个图像-文本对，并且这些描述被仔细地分离成了全局、低频和高频成分。为提高控制能力和一致性，我们在频率感知负面提示的引导下采用了无分类器多分支策略来抑制幻觉效应和语义漂移。 在合成数据集和真实世界基准上的广泛实验表明，DTPSR达到了高感知质量、竞争性的保真度，并且在各种退化场景中具备强大的泛化能力。
URL
https://arxiv.org/abs/2603.07430
