From Zero to Detail: Deconstructing Ultra-High-Definition Image Restoration from Progressive Spectral Perspective

TL;DR

This work tackles UHD image restoration by introducing a progressive spectral perspective that splits restoration into zero-frequency enhancement, low-frequency restoration, and high-frequency refinement. It presents ERR, a tri-branch framework with ZFE (global priors via AAP and GPTB), LFR (mid-resolution coarse content via RSSB), and HFR (high-frequency texture refinement via FW-KAN in the DCT domain). The method uses stage-specific losses $ abla_{ ext{zf}}$, $ abla_{ ext{lf}}$, and $ abla_{ ext{hf}}$, plus per-stage reconstruction losses, to guide learning from global mappings to fine textures. Empirically, ERR achieves state-of-the-art results on four UHD restoration benchmarks with efficient full-resolution inference, and comprehensive ablations validate the necessity and effectiveness of each component.

Abstract

Ultra-high-definition (UHD) image restoration faces significant challenges due to its high resolution, complex content, and intricate details. To cope with these challenges, we analyze the restoration process in depth through a progressive spectral perspective, and deconstruct the complex UHD restoration problem into three progressive stages: zero-frequency enhancement, low-frequency restoration, and high-frequency refinement. Building on this insight, we propose a novel framework, ERR, which comprises three collaborative sub-networks: the zero-frequency enhancer (ZFE), the low-frequency restorer (LFR), and the high-frequency refiner (HFR). Specifically, the ZFE integrates global priors to learn global mapping, while the LFR restores low-frequency information, emphasizing reconstruction of coarse-grained content. Finally, the HFR employs our designed frequency-windowed kolmogorov-arnold networks (FW-KAN) to refine textures and details, producing high-quality image restoration. Our approach significantly outperforms previous UHD methods across various tasks, with extensive ablation studies validating the effectiveness of each component. The code is available at \href{https://github.com/NJU-PCALab/ERR}{here}.

Figures (11)

• Figure 1: Visual comparison with the latest SOTA method.
• Figure 2: Our core motivation. Based on the observations in (a) and (b), we deconstruct the complex UHD restoration problem into three progressive stages: zero-frequency enhancement, low-frequency restoration, and high-frequency refinement.
• Figure 3: Overall framework of our proposed ERR. Building on the insight from progressive spectral perspective, ERR consists of three collaborative sub-networks: the zero-frequency enhancer (ZFE), the low-frequency restorer (LFR), and the high-frequency refiner (HFR).
• Figure 4: The architecture of the Global Perception Transformer Block (GPTB).
• Figure 5: The architecture of the residue state space block (RSSB).