WaveFreqAnchor: Wave-Structural Anchoring and Frequency Correction Diffusion for Training-Free Face Restoration

2026-08-06

AI Summary

This work addresses the degradation of facial structure and fidelity in existing diffusion-based face restoration methods under severe degradation, which stems from insufficient constraints during the reverse diffusion process. To overcome this limitation, the authors propose a training-free restoration framework that jointly enforces geometric consistency and recovers frequency information through wave-structural anchoring and frequency-domain correction. The core innovations include Anchor-Space Wave-Structural Guidance (ASWG) for anisotropic wavelet response consistency, Multi-scale Wavelet-Fourier Injection (MWFI) for phase substitution across scales, and Subband High-Frequency Enhancement (SHE) for detail recovery. Extensive experiments demonstrate that the proposed method significantly outperforms state-of-the-art approaches across various extreme degradation scenarios, achieving high-fidelity and high-definition face restoration.

Abstract

Diffusion-based face restoration that adjusts the sampling trajectory of pre-trained diffusion models has achieved remarkable progress. However, existing approaches provide insufficient constraints during reverse diffusion, causing identity-related structural drift and degraded fidelity under severe degradations. To address this, we propose WaveFreqAnchor, a training-free framework based on Wave-Structural Anchoring and Frequency Correction Diffusion. Specifically, Anchor-Space Wave-Structural Guidance (ASWG) constrains facial structures through anisotropic wave-response consistency, while Multi-scale Wavelet-Fourier Injection (MWFI) aligns the predicted low-frequency subband with the observation by replacing its phase, correcting inconsistencies accumulated during reverse diffusion. For real-world scenes, we further introduce Subband High-Frequency Enhancement (SHE), which performs bounded, spatially masked refinement on the predicted high-frequency subbands to recover fine facial details under unknown compound degradations. Together, these designs effectively preserve facial identity while restoring sharp and realistic facial details. Extensive experiments show that our method consistently outperforms existing methods, achieving high-quality and high-fidelity face restoration.

Problem

• face restoration
• diffusion models
• structural drift
• fidelity degradation
• identity preservation

Innovation

• Wave-Structural Anchoring
• Frequency Correction Diffusion
• Training-Free Face Restoration
• Multi-scale Wavelet-Fourier Injection
• Subband High-Frequency Enhancement