What to Remove, What to Preserve: Dual-Ambiguity Rectification for All-in-One Image Restoration

2026-07-30

AI Summary

This work addresses the semantic and spatial ambiguities inherent in existing unified image restoration methods, where degradation conditions and scene content become entangled in a shared latent space, leading to content distortion and residual artifacts. To explicitly model this dual ambiguity, we propose DAR-Net, which introduces a structured Degradation-Aware Representation (DAR) via simplex constraints and incorporates dedicated Semantic Ambiguity Rectification (SeAR) and Spatial Ambiguity Rectification (SpAR) modules. These modules leverage channel modulation and spatial feature regularization to effectively disentangle degradation and content information within orthogonal response subspaces. Extensive experiments demonstrate that DAR-Net outperforms the strongest baseline by average PSNR gains of 0.14 dB and 0.34 dB under three-class and five-class degradation settings, respectively, while achieving state-of-the-art performance on CDD-11 and WeatherBench benchmarks.

Abstract

All-in-one image restoration aims to handle diverse degradations within a unified framework. Existing methods commonly encode heterogeneous degradation conditions in a shared latent space, where degradation-related cues and scene content can remain entangled. We characterize the resulting challenge as dual ambiguity: semantic ambiguity in channel-wise modulation and spatial ambiguity in restoration responses, which can lead to content corruption and residual artifacts. To mitigate this issue, we propose DAR-Net, a Dual-Ambiguity Rectification Network for all-in-one image restoration. DAR-Net first introduces a Degradation Archetype Representation (DAR) module to construct a structured degradation state through simplex-constrained archetype mixture modeling. Based on this state, a Semantic Ambiguity Rectification (SeAR) module generates degradation-aware prompts to improve channel-wise conditioning in the decoder. A Spatial Ambiguity Rectification (SpAR) module further regularizes degradation-aware and complementary features toward orthogonal response subspaces, reducing spatial interference between removal and preservation cues. Extensive experiments on standard all-in-one restoration benchmarks show that DAR-Net achieves the best overall performance under both three-degradation and five-degradation settings, improving the average PSNR over the strongest competitor by 0.14 dB and 0.34 dB, respectively; it additionally shows superior performance on CDD-11 and WeatherBench.

Problem

• all-in-one image restoration
• dual ambiguity
• semantic ambiguity
• spatial ambiguity
• degradation entanglement

Innovation

• Dual-Ambiguity Rectification
• Degradation Archetype Representation
• Semantic Ambiguity Rectification
• Spatial Ambiguity Rectification
• All-in-One Image Restoration