Preserve More Details: Mitigating Content Drift in Real-World Image Super-Resolution
2026-08-10 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors study how to make low-quality images clearer and more detailed using a technique called Real-ISR. They noticed that when current methods improve these images, the details can get blurry and the meaning of text in the images can change wrongly. To fix this, they created a new model called FSP-Diff that uses two pathways: one to add back fine details and another to keep the text and overall content accurate. Their tests show FSP-Diff works better than previous similar methods at improving image quality without losing important information.
Real-world Image Super-ResolutionStable DiffusionDiffusion ModelsContent DriftSemantic ShiftImage RestorationLow-Quality ImagesTextual SemanticsDetail-Conditioned PathwayImage Fidelity
Authors
Chunxiao Liu, Wei Liu, Anbin Xiong, Erli Meng
Abstract
Real-world image super-resolution (Real-ISR) aims to reconstruct high-quality (HQ) images from low-quality (LQ) inputs subject to diverse real-world degradations. Recent advances have leveraged the LQ inputs and natural image priors learned by Stable Diffusion models to achieve impressive results. However, existing methods often overlook insufficient clarity of LQ inputs inevitably induce content drift in the generated HQ images. This manifests primarily as visual detail degradation and textual semantic shift, severely compromising both fidelity and perceptual quality. To address this challenge, we propose FSP-Diff, a novel one-step diffusion model featuring a dual-pathway architecture. This architecture comprises a Detail-Conditioned Pathway for injecting structured details to recover fine structures, and a Detail-Modulated Semantic Pathway that refines semantic guidance using structured details to mitigate semantic deviations. Extensive experiments on standard Real-ISR benchmarks demonstrate that FSP-Diff surpasses existing one-step diffusion methods in both quantitative and qualitative metrics.