Problem
Research questions and friction points this paper is trying to address.
real-world image restoration
ground-truth supervision
perceptual quality
image quality inconsistency
Beyond Ground-Truth: Leveraging Image Quality Priors for Real-World Image Restoration
🤖 AI Summary
This work addresses the limitation of existing real-world image restoration methods, which rely on ground-truth supervision of inconsistent quality and tend to converge toward outputs with merely average perceptual quality. To overcome this, the authors propose IQPIR, a novel framework that explicitly incorporates no-reference image quality assessment (NR-IQA)-derived quality priors into the restoration process for the first time. By integrating these quality priors with a learned codebook prior through a quality-conditioned Transformer, a dual-branch discrete codebook architecture, and an optimized discrete representation strategy, IQPIR steers the model toward generating perceptually optimal results. The framework is plug-and-play—requiring no modification to the backbone network—and effectively disentangles generic from high-quality-specific features. It outperforms state-of-the-art methods on real image restoration benchmarks and functions as a general-purpose quality-guided module to enhance other restoration models.
📝 Abstract
Real-world image restoration aims to restore high-quality (HQ) images from degraded low-quality (LQ) inputs captured under uncontrolled conditions. Existing methods typically depend on ground-truth (GT) supervision, assuming that GT provides perfect reference quality. However, GT can still contain images with inconsistent perceptual fidelity, causing models to converge to the average quality level of the training data rather than achieving the highest perceptual quality attainable. To address these problems, we propose a novel framework, termed IQPIR, that introduces an Image Quality Prior (IQP)-extracted from pre-trained No-Reference Image Quality Assessment (NR-IQA) models-to guide the restoration process toward perceptually optimal outputs explicitly. Our approach synergistically integrates IQP with a learned codebook prior through three key mechanisms: (1) a quality-conditioned Transformer, where NR-IQA-derived scores serve as conditioning signals to steer the predicted representation toward maximal perceptual quality. This design provides a plug-and-play enhancement compatible with existing restoration architectures without structural modification; and (2) a dual-branch codebook structure, which disentangles common and HQ-specific features, ensuring a comprehensive representation of both generic structural information and quality-sensitive attributes; and (3) a discrete representation-based quality optimization strategy, which mitigates over-optimization effects commonly observed in continuous latent spaces. Extensive experiments on real-world image restoration demonstrate that our method not only surpasses cutting-edge methods but also serves as a generalizable quality-guided enhancement strategy for existing methods. The code is available.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Image Quality Prior
No-Reference IQA
Quality-Conditioned Transformer
Dual-Branch Codebook
Discrete Representation
🔎 Similar Papers
No similar papers found.
