Paper deep dive
What to Remove, What to Preserve: Dual-Ambiguity Rectification for All-in-One Image Restoration
Cencen Liu, Wen Yin, Dongyang Zhang, Dongmin Li, Shan Zhao, Bing Su, Tao He, Jielei Wang, Guoming Lu
Intelligence
Status: succeeded | Model: Gemma-4-26B-A4B | Prompt: intel-v1 | Confidence: 91%
Last extracted: 8/3/2026, 10:20:37 AM
Summary
The paper introduces DAR-Net, a Dual-Ambiguity Rectification Network for all-in-one image restoration. It addresses the entanglement of degradation cues and scene content in shared latent spaces, termed 'dual ambiguity' (semantic and spatial). DAR-Net utilizes a Degradation Archetype Representation (DAR) module to construct structured degradation states, a Semantic Ambiguity Rectification (SeAR) module to generate degradation-aware prompts, and a Spatial Ambiguity Rectification (SpAR) module to enforce orthogonal subspaces for degradation and content features. Experiments demonstrate superior performance on standard benchmarks compared to existing methods.
Entities (13)
Relation Signals (11)
DAR-Net → containsmodule → Degradation Archetype Representation (DAR)
confidence 95% · DAR-Net first introduces a Degradation Archetype Representation (DAR) module
DAR-Net → containsmodule → Semantic Ambiguity Rectification (SeAR)
confidence 95% · a Semantic Ambiguity Rectification (SeAR) module generates degradation-aware prompts
DAR-Net → containsmodule → Spatial Ambiguity Rectification (SpAR)
confidence 95% · A Spatial Ambiguity Rectification (SpAR) module further regularizes
DAR-Net → solves → All-in-One Image Restoration
confidence 95% · DAR-Net is proposed for all-in-one image restoration.
Semantic Ambiguity Rectification (SeAR) → containscomponent → Degradation-Aware Prompt Integrator (DAPI)
confidence 90% · SeAR consists of ... a Degradation-Aware Prompt Integrator (DAPI).
Semantic Ambiguity Rectification (SeAR) → containscomponent → Archetype-Guided Prompt Generator (AGPG)
confidence 90% · SeAR consists of an Archetype-Guided Prompt Generator (AGPG)
DAR-Net → evaluatedon → CDD-11
confidence 90% · it additionally shows superior performance on CDD-11
DAR-Net → evaluatedon → WeatherBench
confidence 90% · it additionally shows superior performance on ... WeatherBench.
Cypher Suggestions (0)
No Cypher suggestions yet.
Abstract
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.
Tags
Links
- Source: https://arxiv.org/abs/2607.28526v1
- Canonical: https://arxiv.org/abs/2607.28526v1
Trouble viewing inline? Open PDF directly →
Full Text
59,235 characters extracted from source content.
Expand or collapse full text
by What to Remove, What to Preserve: Dual-Ambiguity Rectification for All-in-One Image Restoration Cencen Liu University of Electronic Science and Technology of ChinaChengduSichuanChina lcc@std.uestc.edu.cn , Wen Yin University of Electronic Science and Technology of ChinaChengduSichuanChina yinwen1999@std.uestc.edu.cn , Dongyang Zhang University of Electronic Science and Technology of ChinaChengduSichuanChina dyzhang@uestc.edu.cn , Dongmin Li University of Electronic Science and Technology of ChinaChengduSichuanChina lidongmin@std.uestc.edu.cn , Shan Zhao Jiigan TechnologyChengduSichuanChina fly.shanzhao@gmail.com , Bing Su Jiigan TechnologyChengduSichuanChina subing@jiigan.com , Tao He University of Electronic Science and Technology of ChinaChengduSichuanChina tao.he01@hotmail.com , Jielei Wang University of Electronic Science and Technology of ChinaChengduSichuanChina jieleiwang_uestc@163.com and Guoming Lu University of Electronic Science and Technology of ChinaChengduSichuanChina lugm@uestc.edu.cn (2026) 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. All-in-one image restoration, prompt-based restoration, degradation modeling, dual-ambiguity rectification †journalyear: 2026†copyright: c†conference: Proceedings of the 34th ACM International Conference on Multimedia; November 10–14, 2026; Rio de Janeiro, Brazil†booktitle: Proceedings of the 34th ACM International Conference on Multimedia (M ’26), November 10–14, 2026, Rio de Janeiro, Brazil†doi: 10.1145/3767308.3835300†isbn: 979-8-4007-2213-4/2026/11†ccs: Computing methodologies Reconstruction 1. Introduction Image restoration aims to recover clean visual content from degraded observations affected by factors such as noise, haze, and rain conditions. As a front-end step in real-world multimedia content capture and enhancement pipelines, image restoration is also closely related to multimedia content quality. Most existing restoration methods are developed under a predefined degradation setting, and many representative models (Chen et al., 2020, 2022; Zamir et al., 2022) are still deployed in a task-specific manner in practice. Such a paradigm leads to considerable computational and storage overhead and limits generalization in complex environments. To address this limitation, all-in-one image restoration (AIR) has emerged as a unified framework for handling diverse degradations within a single model (Li et al., 2022; Potlapalli et al., 2023; Conde et al., 2024). (a) Content Corruption. (b) Residual Degradation. Figure 1. Typical failure modes caused by degradation-content entanglement. (a) Content Corruption, where faithful image details are mistakenly suppressed. (b) Residual Degradation, where degradation artifacts are retained. Two example panels illustrate failure modes caused by degradation-content entanglement. The first shows content corruption, where faithful image details are suppressed, and the second shows residual degradation, where degradation artifacts remain. Early AIR methods often rely on multi-branch designs, where different degradation types are handled by separate branches or task-specific modules. Although such designs improve degradation specialization, their parameter cost typically scales with the number of degradation types, which limits scalability and makes model deployment increasingly inefficient as restoration scenarios become more diverse (Li et al., 2020; Han et al., 2021). To improve scalability, subsequent studies increasingly adopt shared-backbone conditional restoration, where a unified network is modulated by degradation cues (Li et al., 2022; Potlapalli et al., 2023; Tian et al., 2025). Within this paradigm, different conditioning mechanisms have been explored. Prompt-based methods provide a representative and efficient solution by deriving lightweight conditioning signals from the input and injecting them into a shared backbone, thereby enabling input-adaptive restoration with limited additional parameters (Ai et al., 2024; Jiang et al., 2024; Liu et al., 2025b). In parallel, other alternatives such as MoE-based routing enhance model capacity through dynamic expert selection, allowing different restoration patterns to be handled by different experts when necessary (Yu et al., 2024; Zamfir et al., 2025; Zhang et al., 2024). However, although these methods differ in how they introduce degradation-aware conditioning, they still predominantly rely on shared latent representations within a unified restoration pipeline. As a result, a common difficulty remains insufficiently addressed: degradation-related cues and content-related representations are often encoded in an entangled manner within the shared latent space. Therefore, existing unified restoration models still face difficulty in distinguishing what should be removed from what should be preserved. This challenge is often reflected in two typical failure modes, as shown in Fig. 1: (1) Content Corruption, where faithful image content is mistakenly removed together with degradation, and (2) Residual Degradation, where degradation patterns are not sufficiently suppressed in the restored image. We analyze these failures through two forms of ambiguity. The first is Semantic Ambiguity, namely the channel-wise entanglement between degradation-related cues and content-related representation, which can make degradation-aware modulation less discriminative. The second is Spatial Ambiguity, namely their entanglement in the spatial dimension, which can weaken the spatial selectivity of restoration responses. Motivated by these observations, we propose a Dual-Ambiguity Rectification Network (DAR-Net) for all-in-one image restoration. DAR-Net reduces degradation-content entanglement in both the channel and spatial dimensions through three components. It first employs a Degradation Archetype Representation (DAR) module to construct an archetype-based degradation state, which serves as a degradation prior for subsequent rectification. Based on this degradation state, the Semantic Ambiguity Rectification (SeAR) module alleviates channel-wise entanglement through an Archetype-Guided Prompt Generator (AGPG) and a Degradation-Aware Prompt Integrator (DAPI). Specifically, AGPG first generates a base prompt and then refines it via archetype-guided channel routing conditioned on the degradation state, yielding a degradation-aware prompt. DAPI subsequently injects this rectified prompt into the restoration process for degradation-conditioned feature modulation. In addition, the Spatial Ambiguity Rectification (SpAR) module reduces spatial entanglement via Orthogonal Subspace Rectification (OSR), which treats the rectified prompt as a degradation-aware representation and derives a complementary content-related feature from the latent representation, encouraging the two to occupy orthogonal subspaces. Together, these designs aim to better distinguish removal-related and preservation-related cues in image restoration. The main contributions are summarized as follows: • Dual-ambiguity rectification framework: We propose DAR-Net for AIR, which mitigates degradation-content entanglement in both the channel and spatial dimensions. • Semantic ambiguity rectification: We introduce DAR to model degradation states and SeAR to generate degradation-aware prompts for channel-wise semantic rectification. • Spatial ambiguity rectification: We introduce SpAR with OSR to separate degradation-related and complementary content-related representations in the spatial dimension. • Comprehensive validation and results: Extensive experiments verify that DAR-Net consistently achieves the best overall performance on standard all-in-one restoration benchmarks and generalizes favorably to mixed and real-world degradations. Figure 2. Architecture of the Dual-Ambiguity Rectification Network (DAR-Net). Built upon a hierarchical U-shaped Transformer, DAR-Net mitigates degradation-content entanglement through three components: (1) DAR, which constructs a degradation state via simplex-constrained degradation archetype representation; (2) SeAR, which comprises AGPG to generate a degradation-aware prompt from the degradation state and DAPI to integrate this prompt into the decoder for degradation-conditioned restoration; and (3) SpAR, which is implemented by OSR to regularize the degradation-aware prompt and the corresponding content feature toward orthogonal spatial subspaces under the ℒOSRL_OSR constraint. Block diagram of DAR-Net. A hierarchical U-shaped Transformer processes the input image. DAR constructs a simplex-constrained degradation state, SeAR uses AGPG and DAPI to generate and integrate degradation-aware prompts in the decoder, and SpAR uses OSR to encourage the prompt and content feature to occupy orthogonal spatial subspaces. 2. Related Work 2.1. Task-Specific Image Restoration Traditional image restoration methods are typically designed for a single degradation type, such as denoising (Liang et al., 2021; Chen et al., 2022; Shen et al., 2022), deblurring (Kong et al., 2022; Whang et al., 2021; Cho et al., 2021), deraining (Chen et al., 2023a; Wang et al., 2020; Yi et al., 2021; Xiao et al., 2022), and dehazing (Chen et al., 2023b; Shen et al., 2023; Song et al., 2022; Cai et al., 2016). Their goal is to learn a direct mapping from degraded images to clean images under a predefined degradation setting. With the development of restoration architectures, many methods have gradually moved from heavily customized task-specific designs toward more general restoration backbones. Representative models such as IPT (Chen et al., 2020), SwinIR (Liang et al., 2021), Uformer (Wang et al., 2021), Restormer (Zamir et al., 2022), NAFNet (Chen et al., 2022), and MAXIM (Tu et al., 2022) exemplify this trend by improving restoration quality through stronger feature modeling and broader contextual interaction. Recent image super-resolution methods further explore hybrid Mamba–Transformer modeling to improve efficient long-range interaction (Liu et al., 2025a). While task-specific image restoration methods are effective for individual degradation types, their reliance on separate degradation-specific models leads to poor scalability and limits their applicability in unified real-world restoration settings. 2.2. All-in-one Image Restoration To improve scalability in unified real-world restoration scenarios, all-in-one image restoration aims to handle diverse degradations with a single unified model, making it more suitable for practical settings where the degradation type is unknown or mixed. Existing methods mainly differ in how degradation information is incorporated into a shared restoration pipeline. Representative early directions include degradation representation learning, where AirNet learns contrastive degradation representations (Li et al., 2022); prompt-based conditioning, where PromptIR (Potlapalli et al., 2023), InstructIR (Conde et al., 2024), and UP-Restorer (Liu et al., 2025b) inject learned prompts or instructions into the restoration network; and multimodal guidance, where DA-CLIP (Luo et al., 2024) and MPerceiver (Ai et al., 2024) leverage large-scale vision-language priors for restoration. Subsequent works further improve unified restoration either by strengthening degradation modeling and shared representation learning (Chen et al., 2025; Yang et al., 2024; Jiang et al., 2024; Zhang et al., 2023; Tang et al., 2026) or by introducing degradation-specialized experts to better handle diverse degradation patterns (Zamfir et al., 2025; Zhang et al., 2024; Wang et al., 2025). Despite these different designs, most existing methods still rely on degradation cues to condition, organize, or route shared features within a unified restoration network. In contrast, our method focuses on reducing degradation-content entanglement during feature modulation, so that the model can better distinguish what should be removed from what should be preserved. 3. Methods 3.1. Mathematical Preliminaries Simplices and barycentric coordinates. Let =v1,…,vK⊂ℝDV=\v_1,…,v_K\ ^D with K≤D+1K≤ D+1. If v1,…,vKv_1,…,v_K are affinely independent, then their convex hull (1) conv()=∑k=1Kαkvk|αk≥0,∑k=1Kαk=1conv(V)= \ _k=1^K _kv_k\; |\; _k≥ 0,\; _k=1^K _k=1 \ forms a geometric (K−1)(K-1)-simplex. Equivalently, letting (2) ΔK−1=α∈ℝK|αk≥0,∑k=1Kαk=1, ^K-1= \α ^K\; |\; _k≥ 0,\; _k=1^K _k=1 \, the affine map (3) T:ΔK−1→conv(),T(α)=∑k=1KαkvkT: ^K-1 (V), T(α)= _k=1^K _kv_k is bijective. Hence every point z∈conv()z (V) admits a unique coefficient vector α∈ΔK−1α∈ ^K-1 such that (4) z=∑k=1Kαkvk,z= _k=1^K _kv_k, where α is the barycentric coordinate of z with respect to V. Orthogonal decomposition. Let ℋH be a finite-dimensional inner-product space with inner product ⟨⋅,⋅⟩ ·,· . For any subspace ⊆ℋS , its orthogonal complement is (5) ⟂=y∈ℋ|⟨x,y⟩=0,∀x∈.S = \y \; |\; x,y =0,\ ∀ x \. By the projection theorem, every z∈ℋz admits a unique orthogonal decomposition (6) z=Pz+P⟂z,Pz∈,P⟂z∈⟂,z=P_Sz+P_S z, P_Sz ,\;P_S z , which implies (7) ‖z‖22=‖Pz‖22+‖P⟂z‖22.\|z\|_2^2=\|P_Sz\|_2^2+\|P_S z\|_2^2. More generally, if A⟂BS_A _B, then for any xA∈Ax_A _A and xB∈Bx_B _B, (8) ⟨xA,xB⟩=0,‖xA+xB‖22=‖xA‖22+‖xB‖22. x_A,x_B =0, \|x_A+x_B\|_2^2=\|x_A\|_2^2+\|x_B\|_2^2. In the Euclidean case, let XA,XB∈ℝC×NX_A,X_B ^C× N, and denote their row vectors by xii=1C\x_i\_i=1^C and yjj=1C\y_j\_j=1^C. Then the (i,j)(i,j)-th entry of the cross-Gram matrix satisfies (9) [XAXB⊤]ij=⟨xi,yj⟩.[X_AX_B ]_ij= x_i,y_j . Therefore, the row-generated subspaces of XAX_A and XBX_B are orthogonal if and only if (10) XAXB⊤=C×C.X_AX_B =0_C× C. Moreover, (11) ‖XAXB⊤‖F2=∑i=1C∑j=1C⟨xi,yj⟩2,\|X_AX_B \|_F^2= _i=1^C _j=1^C x_i,y_j ^2, which measures the total pairwise interaction between the two subspaces and vanishes exactly under orthogonality. 3.2. Overview Built upon a hierarchical U-shaped Transformer backbone, DAR-Net mitigates two ambiguities in AIR (Fig. 2). Specifically, we design a rectification pipeline: (1) Degradation Archetype Representation (DAR) (§ 3.3) extracts a global degradation descriptor and maps it to simplex-constrained mixture coefficients to construct a degradation state; (2) Semantic Ambiguity Rectification (SeAR) (§ 3.4) uses this state to rectify prompt channels, yielding a degradation-aware prompt that is further integrated into the decoder for degradation-conditioned feature modulation; and (3) Spatial Ambiguity Rectification (SpAR) (§ 3.5) regularizes the degradation-aware prompt and the corresponding content feature at the deepest decoder stage toward orthogonal subspaces. Finally, the restored image is reconstructed with a global residual connection, and the training objective is given in § 3.6. 3.3. Degradation Archetype Representation The DAR module implements the simplex-constrained parameterization introduced in § 3.1 and provides a structured degradation representation for subsequent rectification. Given an input image Iin∈ℝ3×H×WI_in ^3× H× W, where H and W denote the input height and width, respectively, we first employ a lightweight conditioning network Cond(⋅)Cond(·), implemented by stacked 3×33× 3 convolutional layers, to extract degradation-sensitive features: (12) Fcond=Cond(Iin),Fcond∈ℝC×H×W,F_cond=Cond(I_in), F_cond ^C× H× W, where C is the channel dimension. We then use global average pooling to FcondF_cond to obtain a degradation descriptor ∈ℝCg ^C. This descriptor is projected to a K-dimensional score vector and normalized by a softmax operator to produce the mixture coefficients (13) =Softmax(+),∈ℝK, α=Softmax(Wg+b), α ^K, where ∈ℝK×CW ^K× C, ∈ℝKb ^K, and K denotes the number of degradation archetypes. Here, each entry of α quantifies the contribution of one archetype to the degradation mixture, and thus α serves as the simplex coordinate vector described in § 3.1. Let =[1,2,…,K]∈ℝC×KA=[a_1,a_2,…,a_K] ^C× K denote a learnable degradation archetype matrix, where each column k∈ℝCa_k ^C is an archetypal degradation vector. The degradation state is then constructed as (14) deg=∑k=1Kαkk=,deg∈ℝC,S_deg= _k=1^K _ka_k=A α, _deg ^C, which is a simplex-constrained convex combination of the learned archetypes. As a result, degS_deg lies in the convex hull of the archetypes and serves as the structured degradation representation used in the subsequent rectification modules. 3.4. Semantic Ambiguity Rectification The SeAR mitigates semantic ambiguity, i.e., the channel-wise entanglement between degradation representation and content representation. SeAR consists of an Archetype-Guided Prompt Generator (AGPG) and a Degradation-Aware Prompt Integrator (DAPI). Specifically, SeAR first uses the degradation state degS_deg to generate a degradation-aware prompt, and then integrates this prompt into the stage-wise decoding process. Before the deepest decoder stage, SeAR further derives a content feature by residual decomposition. 3.4.1. Archetype-Guided Prompt Generator (AGPG) Formally, let Fin(l)∈ℝCl×Hl×WlF_in^(l) ^C_l× H_l× W_l denote the input feature before the l-th decoder stage (l∈1,2,3l∈\1,2,3\). AGPG aims to construct a degradation-aware prompt by combining two sources of information: the current-stage feature, which provides input-adaptive prompt cues, and the degradation state degS_deg, which provides structured degradation prior. To this end, we first synthesize a base prompt from a set of M learnable prompt tensors (l)=P1(l),…,PM(l)P^(l)=\P_1^(l),…,P_M^(l)\, where each Pm(l)∈ℝCl×Hl×WlP_m^(l) ^C_l× H_l× W_l. Specifically, we predict an input-dependent mixture weight vector from the globally pooled feature and use it to aggregate the prompt tensors: (15) (l) ^(l) =Softmax(FCp(l)((Fin(l)))), =Softmax\! (FC_p^(l)(G(F_in^(l))) ), Fp(l) F_p^(l) =Conv3×3(l)(∑m=1Mwm(l)Pm(l)). =Conv_3× 3^(l)\! ( _m=1^Mw_m^(l)P_m^(l) ). We then inject the degradation prior by mapping degS_deg to a channel-wise routing vector and using it to rectify the base prompt: (16) ^(l)=σ(FCs(l)(deg)),Fdp(l)=Fp(l)⊙^(l), s^(l)=σ\! (FC_s^(l)(S_deg) ), F_dp^(l)=F_p^(l) s^(l), where ^(l)∈ℝCl×1×1 s^(l) ^C_l× 1× 1 and ⊙ denotes broadcast multiplication over spatial dimensions. In this way, channels that are more consistent with the inferred degradation state are emphasized, while prompt responses unrelated to the current degradation are suppressed. At the deepest decoder stage, we further derive a complementary content-related feature by residual subtraction, (17) Fc(1)=Fin(1)−Fdp(1),F_c^(1)=F_in^(1)-F_dp^(1), and forward the pair (Fdp(1),Fc(1))(F_dp^(1),F_c^(1)) to SpAR for subsequent spatial ambiguity rectification. Table 1. Quantitative comparison on the 3D all-in-one restoration benchmark. We reported PSNR/SSIM. Method Venue Dehazing Deraining Denoising on BSD68 Average SOTS-Outdoor Rain100L σ=15σ=15 σ=25σ=25 σ=50σ=50 Restormer (Zamir et al., 2022) CVPR’22 27.78/0.958 33.78/0.958 33.72/0.930 30.67/0.865 27.63/0.792 30.75/0.901 AirNet (Li et al., 2022) CVPR’22 27.94/0.962 34.90/0.968 33.92/0.933 31.26/0.888 28.00/0.797 31.20/0.910 PromptIR (Potlapalli et al., 2023) NeurIPS’23 30.58/0.974 36.37/0.972 33.98/0.933 31.31/0.888 28.06/0.799 32.06/0.913 InstructIR (Conde et al., 2024) ECCV’24 30.22/0.959 37.98/0.978 34.15/0.933 31.52/0.890 28.30/0.804 32.43/0.913 DiffUIR (Zheng et al., 2024) CVPR’24 30.18/0.973 36.78/0.973 33.94/0.932 31.26/0.887 28.04/0.797 32.04/0.912 AdaIR (Cui et al., 2025) ICLR’25 31.06/0.980 38.64/0.983 34.12/0.935 31.45/0.892 28.19/0.802 32.69/0.918 VLU-Net (Zeng et al., 2025) CVPR’25 30.71/0.980 38.93/0.984 31.13/0.935 31.48/0.892 28.23/0.804 32.10/0.919 MoCE-IR (Zamfir et al., 2025) CVPR’25 31.34/0.979 38.57/0.984 34.11/0.932 31.45/0.888 28.18/0.800 32.73/0.917 DFPIR (Tian et al., 2025) CVPR’25 31.87/0.980 38.65/0.982 34.12/0.935 31.47/0.893 28.25/0.806 32.88/0.919 ClearAIR (Zhang et al., 2026) AAAI’26 31.08/0.981 38.61/0.984 34.18/0.935 31.50/0.891 28.31/0.804 32.74/0.919 MIRAGE (Ren et al., 2026) ICLR’26 31.86/0.981 38.94/0.985 34.12/0.935 31.46/0.891 28.19/0.803 32.91/0.919 DAR-Net (Ours) - 31.93/0.984 39.15/0.986 34.21/0.936 31.58/0.895 28.37/0.808 33.05/0.922 3.4.2. Degradation-Aware Prompt Integrator (DAPI) DAPI injects the degradation-aware prompt into the decoder through channel-wise attention. For a unified formulation, we define (18) Fq(l)=F~dp(1),l=1,Fdp(l),l>1,Fkv(l)=F~c(1),l=1,Fin(l),l>1.F_q^(l)= cases F_dp^(1),&l=1,\\ F_dp^(l),&l>1, cases F_kv^(l)= cases F_c^(1),&l=1,\\ F_in^(l),&l>1. cases Here, the deepest decoder stage uses the SpAR-rectified pair (F~dp(1),F~c(1))( F_dp^(1), F_c^(1)), while later stages directly use the degradation-aware prompt and the current-stage input feature. We then project these inputs into query, key, and value tensors: (19) Q(l)=ΦQ(l)(Fq(l)),K(l)=ΦK(l)(Fkv(l)),V(l)=ΦV(l)(Fkv(l)),Q^(l)= _Q^(l)(F_q^(l)), K^(l)= _K^(l)(F_kv^(l)), V^(l)= _V^(l)(F_kv^(l)), where ΦQ(l) _Q^(l), ΦK(l) _K^(l), and ΦV(l) _V^(l) are three independent projection blocks, each implemented by a 1×11× 1 convolution followed by a depth-wise 3×33× 3 convolution. Let Nl=HlWlN_l=H_lW_l denote the number of spatial locations at the l-th stage. After reshaping Q(l)Q^(l), K(l)K^(l), and V(l)V^(l) to ℝCl×NlR^C_l× N_l, we compute channel-wise attention as (20) A(l)=Softmax(Q(l)K(l)⊤τ),A(l)∈ℝCl×Cl,A^(l)=Softmax\! ( Q^(l)K^(l) τ ), A^(l) ^C_l× C_l, where τ is a learnable temperature parameter and the softmax is applied row-wise. The stage output is then obtained as Fout(l)=Reshape(A(l)V(l)).F_out^(l)=Reshape\! (A^(l)V^(l) ). 3.5. Spatial Ambiguity Rectification SeAR produces a degradation-aware prompt Fdp(1)F_dp^(1) and a complementary content-related feature Fc(1)F_c^(1). Although residual decomposition separates them coarsely, their spatial responses may still remain entangled, leading to spatial ambiguity. To further separate degradation-related and content-related spatial responses before prompt integration, we introduce an Orthogonal Subspace Rectification (OSR) strategy, which encourages the two representations to lie in orthogonal subspaces. Specifically, we first transform Fdp(1)F_dp^(1) and Fc(1)F_c^(1) with two learnable mappings Ωp _p and Ωc _c while preserving their spatial resolution: (21) F~dp(1)=Ωp(Fdp(1)),F~c(1)=Ωc(Fc(1)), F_dp^(1)= _p(F_dp^(1)), F_c^(1)= _c(F_c^(1)), where F~dp(1),F~c(1)∈ℝC1×H1×W1 F_dp^(1), F_c^(1) ^C_1× H_1× W_1. In practice, each mapping is implemented by a 1×11× 1 convolution, a point-wise nonlinearity, and another 1×11× 1 convolution. These learnable transformations allow the model to project the two features into a space where orthogonality can be imposed more effectively. To instantiate the orthogonality constraint, let N1=H1W1N_1=H_1W_1, and let ~dp,k,~c,k∈ℝN1 m_dp,k, m_c,k ^N_1 denote the flattened spatial maps of the k-th channel of F~dp(1) F_dp^(1) and F~c(1) F_c^(1), respectively. Since directly shrinking feature magnitudes could trivially reduce their interaction, we first normalize each channel vector: (22) ^dp,k=~dp,k‖~dp,k‖2+ϵ,^c,k=~c,k‖~c,k‖2+ϵ, m_dp,k= m_dp,k\| m_dp,k\|_2+ε, m_c,k= m_c,k\| m_c,k\|_2+ε, where ϵε is a constant for numerical stability. We then stack the normalized vectors row-wise into Xdp,Xc∈ℝC1×N1X_dp,X_c ^C_1× N_1. In this form, the row spaces of XdpX_dp and XcX_c represent the spatial response subspaces of the degradation-aware and content features, respectively. According to the orthogonal decomposition in § 3.1, two row-generated subspaces are orthogonal if and only if their cross-Gram matrix vanishes, i.e., XdpXc⊤=X_dpX_c =0. We therefore define the OSR loss as (23) ℒOSR=‖XdpXc⊤‖F2=∑i=1C1∑j=1C1⟨^dp,i,^c,j⟩2.L_OSR=\|X_dpX_c \|_F^2= _i=1^C_1 _j=1^C_1 m_dp,i, m_c,j ^2. Minimizing ℒOSRL_OSR suppresses all pairwise inner-product interactions between the channel-wise spatial responses of the two features, thereby encouraging their row-generated subspaces to be orthogonal. The resulting rectified representations F~dp(1) F_dp^(1) and F~c(1) F_c^(1) are then fed into DAPI at the deepest decoder stage. Table 2. Quantitative comparison on the 5D all-in-one restoration benchmark. We reported PSNR/SSIM. Method Venue Dehazing Deraining Denoising Deblurring Low-Light Average SOTS-Outdoor Rain100L BSD68 (σ=25σ=25) GoPro LOL Restormer (Zamir et al., 2022) CVPR’22 24.09/0.927 34.81/0.960 31.49/0.884 27.22/0.829 20.41/0.806 27.60/0.881 AirNet (Li et al., 2022) CVPR’22 21.04/0.884 32.98/0.951 30.91/0.882 24.35/0.781 18.18/0.735 25.49/0.846 PromptIR (Potlapalli et al., 2023) NeurIPS’23 26.54/0.949 36.37/0.970 31.47/0.886 28.71/0.881 22.68/0.832 29.15/0.904 InstructIR (Conde et al., 2024) ECCV’24 27.10/0.956 36.84/0.973 31.40/0.887 29.40/0.886 23.00/0.836 29.55/0.907 DiffUIR (Zheng et al., 2024) CVPR’24 29.47/0.965 35.98/0.968 31.02/0.885 27.50/0.845 22.32/0.826 29.25/0.898 AdaIR (Cui et al., 2025) ICLR’25 30.53/0.978 38.02/0.981 31.35/0.889 28.12/0.858 23.00/0.845 30.20/0.910 VLU-Net (Zeng et al., 2025) CVPR’25 30.84/0.980 38.54/0.982 31.43/0.891 27.46/0.840 22.29/0.833 30.11/0.905 MoCE-IR (Zamfir et al., 2025) CVPR’25 30.48/0.974 38.04/0.982 31.34/0.887 30.05/0.899 23.00/0.852 30.58/0.919 DFPIR (Tian et al., 2025) CVPR’25 31.64/0.979 37.62/0.978 31.29/0.889 28.82/0.873 23.82/0.843 30.64/0.913 ClearAIR (Zhang et al., 2026) AAAI’26 30.12/0.978 38.20/0.982 31.53/0.888 29.67/0.887 22.83/0.846 30.47/0.916 MIRAGE (Ren et al., 2026) ICLR’26 31.45/0.980 38.92/0.982 31.41/0.892 28.10/0.858 23.59/0.858 30.68/0.914 DAR-Net (Ours) - 31.67/0.981 38.34/0.983 31.46/0.892 29.77/0.889 23.86/0.860 31.02/0.921 Table 3. Quantitative Comparison on CDD-11 (Guo et al., 2024) Dataset. We reported PSNR/SSIM metrics. Method Single Double Triple Average L H R S L+H L+R L+S H+R H+S L+H+R L+H+S AirNet (Li et al., 2022) 24.83/0.778 24.21/0.951 26.55/0.891 26.79/0.919 23.23/0.779 22.82/0.710 23.29/0.723 22.21/0.868 23.29/0.901 21.80/0.708 22.24/0.725 23.75/0.814 PromptIR (Potlapalli et al., 2023) 26.32/0.805 26.10/0.969 31.56/0.946 31.53/0.960 24.49/0.789 25.05/0.771 24.51/0.761 24.54/0.924 23.70/0.925 23.74/0.752 23.33/0.747 25.90/0.850 WeatherDiff (Özdenizci and Legenstein, 2022) 23.58/0.763 21.99/0.904 24.85/0.885 24.80/0.888 21.83/0.756 22.69/0.730 22.12/0.707 21.25/0.868 21.99/0.868 21.23/0.716 21.04/0.698 22.49/0.799 WGWS-Net (Zhu et al., 2023) 24.39/0.774 27.90/0.982 33.15/0.964 34.43/0.973 24.27/0.800 25.06/0.772 24.60/0.765 27.23/0.955 27.65/0.960 23.90/0.772 23.97/0.771 26.96/0.863 OneRestore (Guo et al., 2024) 26.48/0.826 32.52/0.990 33.40/0.964 34.31/0.973 25.79/0.822 25.58/0.799 25.19/0.789 29.99/0.957 30.21/0.964 24.78/0.788 24.90/0.791 28.47/0.878 AdaIR (Cui et al., 2025) 26.88/0.821 31.60/0.987 33.84/0.962 34.65/0.974 25.69/0.811 25.90/0.793 25.69/0.783 29.38/0.955 28.95/0.961 24.82/0.778 25.04/0.778 28.40/0.873 MoCE-IR (Zamfir et al., 2025) 27.26/0.824 32.66/0.990 34.31/0.970 35.91/0.980 26.24/0.817 26.25/0.800 26.04/0.793 29.93/0.964 30.19/0.970 25.41/0.789 25.39/0.790 29.05/0.881 DAR-Net (Ours) 27.54/0.834 34.21/0.991 34.96/0.972 36.43/0.981 26.74/0.830 26.63/0.812 26.58/0.806 31.25/0.968 31.28/0.971 25.53/0.801 25.83/0.799 29.73/0.888 Table 4. Quantitative Comparison of different methods on WeatherBench (Guan et al., 2025) Dataset. Method Dehazing Deraining Desnowing Average PSNR↑ SSIM↑ LPIPS↓ FID↓ PSNR↑ SSIM↑ LPIPS↓ FID↓ PSNR↑ SSIM↑ LPIPS↓ FID↓ PSNR↑ SSIM↑ LPIPS↓ FID↓ AirNet (Li et al., 2022) 19.27 0.645 0.3829 134.09 31.56 0.912 0.2236 125.54 20.58 0.737 0.2912 138.57 23.80 0.764 0.2992 132.73 TransWeather (Valanarasu et al., 2021) 18.13 0.621 0.3970 123.21 28.59 0.880 0.2638 149.66 24.06 0.754 0.2250 102.99 23.59 0.752 0.2953 125.29 PromptIR (Potlapalli et al., 2023) 19.50 0.658 0.3751 113.55 32.51 0.915 0.1980 111.69 26.35 0.804 0.1951 84.12 26.12 0.792 0.2561 103.12 WGWS-Net (Zhu et al., 2023) 11.78 0.532 0.5351 152.76 34.77 0.939 0.1168 60.99 19.39 0.721 0.2481 128.56 21.98 0.731 0.3000 114.10 Histoformer (Sun et al., 2024) 15.82 0.597 0.4371 128.34 28.87 0.876 0.2785 152.42 23.88 0.769 0.2252 105.82 22.86 0.747 0.3136 128.86 AdaIR (Cui et al., 2025) 21.39 0.680 0.3506 110.07 32.81 0.918 0.1916 109.41 26.87 0.806 0.1790 73.48 27.02 0.801 0.2404 97.65 DiffUIR (Zheng et al., 2024) 20.96 0.695 0.3550 127.54 33.78 0.931 0.1720 86.96 27.87 0.844 0.1619 68.99 27.54 0.823 0.2296 94.50 DAR-Net (Ours) 23.44 0.732 0.3257 108.35 35.48 0.941 0.1663 82.65 29.37 0.872 0.1569 65.28 29.43 0.848 0.2163 85.43 3.6. Training Objective DAR-Net is trained with a pixel-wise reconstruction loss and the orthogonality regularization introduced in § 3.5. The overall loss is (24) ℒtotal=ℒrec+λℒOSR,L_total=L_rec+λ\,L_OSR, where λ is a balancing coefficient. We adopt the L1 loss between the restored image IoutI_out and the ground-truth image IgtI_gt as the reconstruction loss: (25) ℒrec=1N∑i=1N|Iout(i)−Igt(i)|,L_rec= 1N _i=1^N |I_out^(i)-I_gt^(i) |, where N denotes the total number of image elements. The term ℒOSRL_OSR, defined in Eq. (23), regularizes the degradation-aware prompt and the corresponding content feature at the deepest decoder stage by encouraging their spatial response subspaces to be orthogonal. 4. Experiments We evaluate DAR-Net under both three-degradation (3D) and five-degradation (5D) all-in-one restoration settings. Beyond standard evaluation, we further assess its generalization ability on mixed degradations, and real-world images. We compare DAR-Net with representative restoration methods, including Restormer (Zamir et al., 2022), AirNet (Li et al., 2022), PromptIR (Potlapalli et al., 2023), InstructIR (Conde et al., 2024), DiffUIR (Zheng et al., 2024), AdaIR (Cui et al., 2025), VLU-Net (Zeng et al., 2025), MoCE-IR (Zamfir et al., 2025), DFPIR (Tian et al., 2025), ClearAIR (Zhang et al., 2026), MIRAGE (Ren et al., 2026), WeatherDiff (Özdenizci and Legenstein, 2022), WGWS-Net (Zhu et al., 2023), OneRestore (Guo et al., 2024), TransWeather (Valanarasu et al., 2021) and Histoformer (Sun et al., 2024). We use PSNR and SSIM (Wang et al., 2004) for pixel-wise fidelity evaluation, and LPIPS (Zhang et al., 2018) and FID (Heusel et al., 2017) for perceptual quality assessment. Unless otherwise specified, the results of the compared methods are taken from their original papers or from the survey (Jiang et al., 2025). The best and second-best results are highlighted in bold and underlined, respectively. Figure 3. Qualitative comparison of 3D all-in-one restoration results. A grid of qualitative examples compares image restoration results from multiple methods on the three-degradation all-in-one restoration benchmark. 4.1. Experimental Settings Datasets. For the 3D setting, we train on BSD400 (Arbelaez et al., 2010) and WED (Ma et al., 2016), and evaluate denoising on BSD68 (Martin et al., 2001) with Gaussian noise levels σ∈15,25,50σ∈\15,25,50\. Rain100L (Yang et al., 2017) and SOTS (Li et al., 2018) are used for deraining and dehazing, respectively. For the 5D setting, we further include GoPro (Nah et al., 2017) for deblurring and LOL (Wei et al., 2018) for low-light enhancement. For mixed-degradation evaluation, we use CDD-11 (Guo et al., 2024). For real-world evaluation, we adopt the WeatherBench (Guan et al., 2025). Implementation details. Our model is built on a hierarchical U-shaped Transformer backbone. We use a 4-level encoder-decoder architecture with [4, 6, 6, 8] Transformer blocks from level-1 to level-4. We optimize the network using AdamW with an initial learning rate of 2×10−42× 10^-4 and β1=0.9 _1=0.9, β2=0.99 _2=0.99. The learning rate is decayed to 1×10−81× 10^-8 using cosine annealing with five cycles. The model is trained for 450K and 650K iterations under the 3D and 5D settings, respectively, with a batch size of 32. During training, input images are randomly cropped into 128×128128× 128 patches and augmented by random flipping and rotation. The number of degradation archetypes K=16K=16, the temperature parameter τ=0.07τ=0.07, and the loss weight λ=0.1λ=0.1. All experiments are implemented in PyTorch and conducted on 2 NVIDIA A800 GPUs. 4.2. Main Results Three-Degradation Evaluation. As shown in Tab. 1, DAR-Net achieves the best overall performance under the 3D setting, with an average PSNR/SSIM of 33.05/0.922. It consistently ranks first on dehazing, deraining, and all three denoising levels, demonstrating strong and balanced restoration performance across different degradation types. Compared with the second-best method, DAR-Net improves the average PSNR by 0.14 dB. These results indicate that DAR-Net can more effectively handle degradation-content entanglement in the all-in-one restoration setting, leading to both stronger degradation removal and better content preservation. Five-Degradation Evaluation. DAR-Net achieves the best overall performance under the 5D setting, with an average PSNR/SSIM of 31.02/0.921 (Tab. 2). Compared with the second-best method, DAR-Net improves the average PSNR by 0.34 dB. Although it is not the best-performing method on every task, DAR-Net achieves the best results on dehazing and low-light enhancement while remaining competitive on deraining, denoising, and deblurring. These results indicate that DAR-Net maintains a strong overall balance across diverse degradation types in the more challenging 5D setting. Table 5. Ablation on the key components of DAR-Net. DAR SeAR SpAR PSNR SSIM ✗ ✗ ✗ 29.15 0.904 ✓ ✗ ✗ 29.22 0.905 ✓ ✓ ✗ 30.65 0.916 ✓ ✓ ✓ 31.02 0.921 Table 6. Ablation on content feature construction. Method PSNR SSIM No decomposition 30.78 0.917 Gated suppression 30.84 0.918 Residual subtraction 31.02 0.921 Mixed-degradation Evaluation. As shown in Tab. 3, DAR-Net achieves the best results on all CDD-11 (Guo et al., 2024) subsets, covering single, double, and triple degradations. It obtains the highest average PSNR/SSIM of 29.73/0.888, surpassing the second-best method by 0.68 dB in PSNR and 0.007 in SSIM. The consistent gains across increasingly complex degradation combinations verify the effectiveness of DAR-Net for mixed-degradation restoration. Real-world Evaluation. Tab. 4 reports the quantitative comparison on the WeatherBench (Guan et al., 2025) dataset. DAR-Net achieves the best overall performance, with particularly clear advantages on dehazing and desnowing. On deraining, DAR-Net also attains the best PSNR and SSIM, while remaining competitive in LPIPS and FID. These results demonstrate that DAR-Net generalizes effectively to real-world weather degradations and yields restoration results with improved fidelity and perceptual quality. Qualitative Results. Fig. 3 presents qualitative results under the 3D setting. DAR-Net removes degradations more thoroughly across diverse restoration tasks while better preserving natural image structures. For example, in the deraining case, our result is free of visible rain-streak residue, whereas competing methods still retain noticeable artifacts. In the denoising example with σ=25σ=25, DAR-Net suppresses noise effectively without mistakenly removing the cloud structures in the sky. 4.3. Ablation Study Effect of Key Components. Tab. 6 reports only the average results under the 5D setting for clarity. The full DAR-Net achieves the best performance, validating the effectiveness of the overall design and the complementarity of its three components. DAR provides a structured degradation prior, while SeAR yields more substantial gains by alleviating channel-wise semantic ambiguity. SpAR further improves the performance, and the combination of all three components leads to the best overall result. Effect of Content Feature Construction. We analyze how to construct the content feature FcF_c in SpAR while keeping DAR, SeAR, and SpAR enabled. Specifically, we compare three variants: no decomposition (Fc=FinF_c=F_in), gated suppression (Fc=Fin⊙(1−σ(Fdp))F_c=F_in (1-σ(F_dp))), and residual subtraction (Fc=Fin−FdpF_c=F_in-F_dp). Here, FinF_in denotes the mixed input feature and FdpF_dp denotes the degradation-related feature in § 3.5. As shown in Tab. 6, the residual formulation achieves the best performance, suggesting that explicitly subtracting degradation-related information is more effective for isolating content. Effect of SpAR Placement. Applying SpAR at the deepest decoder stage yields the best performance; detailed placement results are provided in the supplementary material. Figure 4. Visualization of the average simplex coefficients learned by DAR for different degradation types. Distinct degradations exhibit different archetype activation patterns, while related degradations still share partial archetypes. A visualization compares the average simplex coefficients learned by DAR across degradation types. Different degradation types activate distinct archetype patterns, while related degradations share some archetype activations. Figure 5. Intra-/inter-class similarity distributions of prompt features before and after SeAR. SeAR increases intra-class similarity while decreasing inter-class similarity. Similarity distributions of prompt features before and after SeAR show that SeAR shifts intra-class similarity upward and inter-class similarity downward. 4.4. Analysis Analysis of DAR. Fig. 4 shows that different degradations activate distinct archetype combinations, while related degradations share partial archetypes, indicating structured yet transferable degradation representations. Analysis of the archetype number K is provided in the supplementary material. Analysis of SeAR. As shown in Fig. 5, SeAR increases intra-class prompt similarity from 0.63 to 0.74 and decreases inter-class similarity from 0.58 to 0.42, demonstrating improved degradation discrimination. Analysis of SpAR. As shown in Fig. 6, removing FdpF_dp leaves residual degradations, whereas removing FcF_c damages structural content, confirming their complementary roles in degradation removal and content preservation. Figure 6. Analysis of SpAR. F1, F2, and F3 denote the results without the degradation-related feature, without the content-related feature, and with the full model, respectively. The difference maps indicate distinct roles of the two features in degradation removal and content preservation. SpAR analysis compares F1 without the degradation-related feature, F2 without the content-related feature, and F3 from the full model. Difference maps show that the two features play distinct roles in removing degradation and preserving content. Model Complexity and Efficiency. As shown in Tab. 7, DAR-Net has 35.5M parameters and 771G FLOPs, which are comparable to existing methods. Despite slightly higher complexity than some lightweight baselines, DAR-Net still achieves competitive inference efficiency, with lower latency than PromptIR, AdaIR, and DFPIR. Compared with MoCE-IR, DAR-Net incurs moderate additional overhead while providing stronger restoration performance, demonstrating a favorable efficiency-performance trade-off. Table 7. Comparison of model complexity and inference efficiency. FLOPs and latency are measured on an input image of size 720×480720× 480 on a single NVIDIA A800 GPU. Method PromptIR AdaIR MoCE-IR DFPIR DAR-Net Params. 34.1M 28.8M 25.4M 31M + 63M 35.5M FLOPs 752G 786G 493G 885G 771G Latency 187ms 239ms 161ms 193ms 165ms CPU Memory 4454M 4336M 4389M 4369M 4200M GPU Memory 3324M 3117M 1445M 3459M 3162M 5. Conclusion In this paper, we presented DAR-Net, a dual-ambiguity rectification network for all-in-one image restoration. We identify that existing unified restoration methods often suffer from semantic ambiguity in channel-wise representations and spatial ambiguity in spatial responses. To address this, we introduced DAR to learn structured degradation states, SeAR to improve channel-wise degradation discrimination, and SpAR to reduce spatial entanglement between degradation and content. Extensive experiments demonstrate that DAR-Net achieves strong and balanced restoration performance across diverse degradations. These results suggest that explicitly modeling what should be removed and what should be preserved is an effective direction for unified image restoration. Acknowledgements.This research was partially supported by the National Natural Science Foundation of China (NSFC) (62306064) and the Sichuan Science and Technology Program (granted No. 2024ZDZX0011, No. 2026NSFSC1482 and No. 2025ZHCG0002). References Y. Ai, H. Huang, X. Zhou, J. Wang, and R. He (2024) Multimodal prompt perceiver: empower adaptiveness generalizability and fidelity for all-in-one image restoration. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, p. 25432–25444. Cited by: §1, §2.2. P. Arbelaez, M. Maire, C. Fowlkes, and J. Malik (2010) Contour detection and hierarchical image segmentation. IEEE transactions on pattern analysis and machine intelligence 33 (5), p. 898–916. Cited by: §4.1. B. Cai, X. Xu, K. Jia, C. Qing, and D. Tao (2016) DehazeNet: an end-to-end system for single image haze removal. IEEE Transactions on Image Processing 25, p. 5187–5198. Cited by: §2.1. H. Chen, Y. Wang, T. Guo, C. Xu, Y. Deng, Z. Liu, S. Ma, C. Xu, C. Xu, and W. Gao (2020) Pre-trained image processing transformer. 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), p. 12294–12305. Cited by: §1, §2.1. I. Chen, W. Chen, Y. Liu, Y. Chiang, S. Kuo, and M. Yang (2025) UniRestore: unified perceptual and task-oriented image restoration model using diffusion prior. 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), p. 17969–17979. Cited by: §2.2. L. Chen, X. Chu, X. Zhang, and J. Sun (2022) Simple baselines for image restoration. In Computer Vision – ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part VII, Berlin, Heidelberg, p. 17–33. External Links: ISBN 978-3-031-20070-0 Cited by: §1, §2.1. X. Chen, H. Li, M. Li, and J. Pan (2023a) Learning a sparse transformer network for effective image deraining. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), p. 5896–5905. Cited by: §2.1. Z. Chen, Z. He, and Z. Lu (2023b) DEA-net: single image dehazing based on detail-enhanced convolution and content-guided attention. IEEE Transactions on Image Processing 33, p. 1002–1015. Cited by: §2.1. S. Cho, S. Ji, J. Hong, S. Jung, and S. Ko (2021) Rethinking coarse-to-fine approach in single image deblurring. 2021 IEEE/CVF International Conference on Computer Vision (ICCV), p. 4621–4630. Cited by: §2.1. M. V. Conde, G. Geigle, and R. Timofte (2024) InstructIR: high-quality image restoration following human instructions. In Computer Vision – ECCV 2024: 18th European Conference, Milan, Italy, September 29–October 4, 2024, Proceedings, Part XXXVI, Berlin, Heidelberg, p. 1–21. External Links: ISBN 978-3-031-72763-4 Cited by: §1, §2.2, Table 1, Table 2, §4. Y. Cui, S. W. Zamir, S. Khan, A. Knoll, M. Shah, and F. S. Khan (2025) Adair: adaptive all-in-one image restoration via frequency mining and modulation. In 13th international conference on learning representations, ICLR 2025, p. 57335–57356. Cited by: Table 1, Table 2, Table 3, Table 4, §4. Q. Guan, Q. Yang, X. Chen, T. Song, G. Jin, and J. Jin (2025) WeatherBench: a real-world benchmark dataset for all-in-one adverse weather image restoration. Proceedings of the 33rd ACM International Conference on Multimedia. Cited by: Table 4, Table 4, §4.1, §4.2. Y. Guo, Y. Gao, Y. Lu, H. Zhu, R. W. Liu, and S. He (2024) Onerestore: a universal restoration framework for composite degradation. In European conference on computer vision, p. 255–272. Cited by: Table 3, Table 3, Table 3, §4.1, §4.2, §4. J. Han, W. Li, P. Fang, C. Sun, J. Hong, M. A. Armin, L. Petersson, and H. Li (2021) Blind image decomposition. In European Conference on Computer Vision, Cited by: §1. M. Heusel, H. Ramsauer, T. Unterthiner, B. Nessler, and S. Hochreiter (2017) Gans trained by a two time-scale update rule converge to a local nash equilibrium. Advances in neural information processing systems 30. Cited by: §4. J. Jiang, Z. Zuo, G. Wu, K. Jiang, and X. Liu (2025) A survey on all-in-one image restoration: taxonomy, evaluation and future trends. IEEE Transactions on Pattern Analysis and Machine Intelligence. Cited by: §4. Y. Jiang, Z. Zhang, T. Xue, and J. Gu (2024) Autodir: automatic all-in-one image restoration with latent diffusion. In European Conference on Computer Vision, p. 340–359. Cited by: §1, §2.2. L. Kong, J. Dong, M. Li, J. Ge, and J. Pan (2022) Efficient frequency domain-based transformers for high-quality image deblurring. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), p. 5886–5895. Cited by: §2.1. B. Li, W. Ren, D. Fu, D. Tao, D. Feng, W. Zeng, and Z. Wang (2018) Benchmarking single-image dehazing and beyond. IEEE transactions on image processing 28 (1), p. 492–505. Cited by: §4.1. B. Li, X. Liu, P. Hu, Z. Wu, J. Lv, and X. Peng (2022) All-in-one image restoration for unknown corruption. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), p. 17431–17441. Cited by: §1, §1, §2.2, Table 1, Table 2, Table 3, Table 4, §4. R. Li, R. T. Tan, and L. F. Cheong (2020) All in one bad weather removal using architectural search. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), p. 3172–3182. Cited by: §1. J. Liang, J. Cao, G. Sun, K. Zhang, L. Van Gool, and R. Timofte (2021) SwinIR: image restoration using swin transformer. 2021 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW), p. 1833–1844. Cited by: §2.1. C. Liu, D. Zhang, G. Lu, W. Yin, J. Wang, and G. Luo (2025a) SRMamba-T: exploring the hybrid Mamba–Transformer network for single image super-resolution. Neurocomputing 624, p. 129488. Cited by: §2.1. M. Liu, W. Yang, J. Luo, and J. Liu (2025b) Up-restorer: when unrolling meets prompts for unified image restoration. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 39, p. 5513–5522. Cited by: §1, §2.2. Z. Luo, F. K. Gustafsson, Z. Zhao, J. Sjölund, and T. B. Schön (2024) Controlling vision-language models for multi-task image restoration. In The Twelfth International Conference on Learning Representations, Vienna, Austria, May 7, 2024, Cited by: §2.2. K. Ma, Z. Duanmu, Q. Wu, Z. Wang, H. Yong, H. Li, and L. Zhang (2016) Waterloo exploration database: new challenges for image quality assessment models. IEEE Transactions on Image Processing 26 (2), p. 1004–1016. Cited by: §4.1. D. Martin, C. Fowlkes, D. Tal, and J. Malik (2001) A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics. In Proceedings eighth IEEE international conference on computer vision. ICCV 2001, Vol. 2, p. 416–423. Cited by: §4.1. S. Nah, T. Hyun Kim, and K. Mu Lee (2017) Deep multi-scale convolutional neural network for dynamic scene deblurring. In Proceedings of the IEEE conference on computer vision and pattern recognition, p. 3883–3891. Cited by: §4.1. O. Özdenizci and R. A. Legenstein (2022) Restoring vision in adverse weather conditions with patch-based denoising diffusion models. IEEE Transactions on Pattern Analysis and Machine Intelligence 45, p. 10346–10357. Cited by: Table 3, §4. V. Potlapalli, S. W. Zamir, S. Khan, and F. S. Khan (2023) PromptIR: prompting for all-in-one blind image restoration. In Proceedings of the 37th International Conference on Neural Information Processing Systems, NIPS ’23, Red Hook, NY, USA. Cited by: §1, §1, §2.2, Table 1, Table 2, Table 3, Table 4, §4. B. Ren, Y. Li, X. Zheng, Y. Fu, D. P. Paudel, H. Liu, M. Yang, L. V. Gool, and N. Sebe (2026) Efficient degradation-agnostic image restoration via channel-wise functional decomposition and manifold regularization. In The Fourteenth International Conference on Learning Representations, Cited by: Table 1, Table 2, §4. H. Shen, Z. Zhao, Y. Zhang, and Z. Zhang (2023) Mutual information-driven triple interaction network for efficient image dehazing. Proceedings of the 31st ACM International Conference on Multimedia. Cited by: §2.1. H. Shen, Z. Zhao, and W. Zhang (2022) Adaptive dynamic filtering network for image denoising. In AAAI Conference on Artificial Intelligence, Cited by: §2.1. Y. Song, Z. He, H. Qian, and X. Du (2022) Vision transformers for single image dehazing. IEEE Transactions on Image Processing 32, p. 1927–1941. Cited by: §2.1. S. Sun, W. Ren, X. Gao, R. Wang, and X. Cao (2024) Restoring images in adverse weather conditions via histogram transformer. In European Conference on Computer Vision, Cited by: Table 4, §4. X. Tang, X. He, J. Xu, X. Gu, and J. Sun (2026) Learning continuous wasserstein barycenter space for generalized all-in-one image restoration. IEEE Transactions on Pattern Analysis and Machine Intelligence. Cited by: §2.2. X. Tian, X. Liao, X. Liu, M. Li, and C. Ren (2025) Degradation-aware feature perturbation for all-in-one image restoration. 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), p. 28165–28175. Cited by: §1, Table 1, Table 2, §4. Z. Tu, H. Talebi, H. Zhang, F. Yang, P. Milanfar, A. C. Bovik, and Y. Li (2022) MAXIM: multi-axis mlp for image processing. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), p. 5759–5770. Cited by: §2.1. J. M. J. Valanarasu, R. Yasarla, and V. M. Patel (2021) TransWeather: transformer-based restoration of images degraded by adverse weather conditions. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), p. 2343–2353. Cited by: Table 4, §4. C. Wang, Y. Wu, Z. Su, and J. Chen (2020) Joint self-attention and scale-aggregation for self-calibrated deraining network. Proceedings of the 28th ACM International Conference on Multimedia. Cited by: §2.1. Y. Wang, Y. Li, Z. Zheng, X. Zhang, and M. Wei (2025) M2Restore: mixture-of-experts-based mamba-cnn fusion framework for all-in-one image restoration. IEEE Transactions on Image Processing 34, p. 8086–8100. Cited by: §2.2. Z. Wang, X. Cun, J. Bao, and J. Liu (2021) Uformer: a general u-shaped transformer for image restoration. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), p. 17662–17672. Cited by: §2.1. Z. Wang, A.C. Bovik, H.R. Sheikh, and E.P. Simoncelli (2004) Image quality assessment: from error visibility to structural similarity. IEEE Transactions on Image Processing 13 (4), p. 600–612. Cited by: §4. C. Wei, W. Wang, W. Yang, and J. Liu (2018) Deep retinex decomposition for low-light enhancement. In British Machine Vision Conference 2018, p. 155. Cited by: §4.1. J. Whang, M. Delbracio, H. Talebi, C. Saharia, A. G. Dimakis, and P. Milanfar (2021) Deblurring via stochastic refinement. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), p. 16272–16282. Cited by: §2.1. J. Xiao, X. Fu, A. Liu, F. Wu, and Z. Zha (2022) Image de-raining transformer. IEEE Transactions on Pattern Analysis and Machine Intelligence 45, p. 12978–12995. Cited by: §2.1. H. Yang, L. Pan, Y. Yang, and W. Liang (2024) Language-driven all-in-one adverse weather removal. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, p. 24902–24912. Cited by: §2.2. W. Yang, R. T. Tan, J. Feng, J. Liu, Z. Guo, and S. Yan (2017) Deep joint rain detection and removal from a single image. In Proceedings of the IEEE conference on computer vision and pattern recognition, p. 1357–1366. Cited by: §4.1. Q. Yi, J. Li, Q. Dai, F. Fang, G. Zhang, and T. Zeng (2021) Structure-preserving deraining with residue channel prior guidance. 2021 IEEE/CVF International Conference on Computer Vision (ICCV), p. 4218–4227. Cited by: §2.1. X. Yu, S. Zhou, H. Li, and L. Zhu (2024) Multi-expert adaptive selection: task-balancing for all-in-one image restoration. IEEE Transactions on Circuits and Systems for Video Technology 35, p. 4619–4634. Cited by: §1. E. Zamfir, Z. Wu, N. Mehta, Y. Tan, D. P. Paudel, Y. Zhang, and R. Timofte (2025) Complexity experts are task-discriminative learners for any image restoration. 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), p. 12753–12763. Cited by: §1, §2.2, Table 1, Table 2, Table 3, §4. S. W. Zamir, A. Arora, S. H. Khan, M. Hayat, F. S. Khan, and M. Yang (2022) Restormer: efficient transformer for high-resolution image restoration. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), p. 5718–5729. Cited by: §1, §2.1, Table 1, Table 2, §4. H. Zeng, X. Wang, Y. Chen, J. Su, and J. Liu (2025) Vision-language gradient descent-driven all-in-one deep unfolding networks. 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), p. 7524–7533. Cited by: Table 1, Table 2, §4. J. Zhang, J. Huang, M. Yao, Z. Yang, H. Yu, M. Zhou, and F. Zhao (2023) Ingredient-oriented multi-degradation learning for image restoration. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), p. 5825–5835. Cited by: §2.2. R. Zhang, P. Isola, A. A. Efros, E. Shechtman, and O. Wang (2018) The unreasonable effectiveness of deep features as a perceptual metric. In Proceedings of the IEEE conference on computer vision and pattern recognition, p. 586–595. Cited by: §4. R. Zhang, Y. Luo, J. Liu, H. Yang, Z. Dong, D. Gudovskiy, T. Okuno, Y. Nakata, K. Keutzer, Y. Du, et al. (2024) Efficient deweahter mixture-of-experts with uncertainty-aware feature-wise linear modulation. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 38, p. 16812–16820. Cited by: §1, §2.2. X. Zhang, H. Zhang, G. Wang, Q. Zhang, L. Zhang, and B. Du (2026) ClearAIR: a human-visual-perception-inspired all-in-one image restoration. In Proceedings of the AAAI Conference on Artificial Intelligence, Cited by: Table 1, Table 2, §4. D. Zheng, X. Wu, S. Yang, J. Zhang, J. Hu, and W. Zheng (2024) Selective hourglass mapping for universal image restoration based on diffusion model. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, p. 25445–25455. Cited by: Table 1, Table 2, Table 4, §4. Y. Zhu, T. Wang, X. Fu, X. Yang, X. Guo, J. Dai, Y. Qiao, and X. Hu (2023) Learning weather-general and weather-specific features for image restoration under multiple adverse weather conditions. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), p. 21747–21758. Cited by: Table 3, Table 4, §4.