Degradation Frequency Curve: An Explicit Frequency-Quantified Representation for All-in-One Image Restoration

2026-05-17

AI Summary

This work addresses the challenge in blind image restoration where degradation processes are often implicit and difficult to explicitly quantify or manipulate under mixed, composite, or unseen degradations. The authors propose the Degradation Frequency Curve (DFC), which, for the first time, models degradation as a quantifiable representation in the frequency domain—specifically, the per-band residual-to-degraded-image energy ratio. Building upon this, they introduce a multi-scale token-conditioned restoration framework, termed DFC-IR. By leveraging spectral token decomposition, DFC-IR enables structured representation of degradation effects and facilitates the extraction of reusable priors, thereby overcoming the limitations of conventional implicit mapping strategies. Experiments demonstrate that DFC-IR achieves state-of-the-art performance across standard, composite, unseen, and real-world degradation benchmarks, significantly enhancing generalization in complex degradation scenarios.

Abstract

A fundamental difficulty in all-in-one blind image restoration is that degradation is usually treated as an implicit factor hidden in degraded-to-clean mapping, rather than as an explicit object that can be measured and manipulated. This limitation becomes more pronounced under mixed, compound, or unseen degradation conditions, where degradation effects are hard to assign to predefined labels or task-specific parameters. We propose the Degradation Frequency Curve (DFC), a structured spectral representation that quantifies degradation responses by measuring band-wise residual-to-degraded energy ratios in the frequency domain. DFC converts visually entangled and hard-to-describe degradation effects into a measurable degradation coordinate space. Moreover, DFC can be adaptively decomposed into band-wise spectral tokens, allowing local degradation responses to be represented as reusable restoration priors. Based on this representation, we develop the DFC-guided Image Restorer (DFC-IR), a token-conditioned multi-scale framework that progressively estimates DFCs from intermediate restorations and uses the resulting spectral tokens to guide degradation-aware restoration in a coarse-to-fine manner. Extensive experiments on standard, composite, unseen, and real-world degradation benchmarks show that DFC provides an effective representation basis for all-in-one restoration, leading to state-of-the-art performance and improved generalization under complex degradation profiles.

Problem

• blind image restoration
• degradation representation
• frequency domain
• all-in-one restoration
• unseen degradation

Innovation

• Degradation Frequency Curve
• frequency-domain representation
• all-in-one image restoration
• spectral tokens
• blind image restoration