FluidRain: Incompressible Rain Flow as an Attention Bias for Loop-in-Loop Video Deraining

Pu Wang; Yongcong Wang; Wenhao Li; Xiang Chen; Guangwei Gao; Jinshan Pan; Siyuan Yao; Shujun Fu

cs.CVSep 24, 2026

FluidRain: Incompressible Rain Flow as an Attention Bias for Loop-in-Loop Video Deraining

Authors: Pu Wang, Yongcong Wang, Wenhao Li, Xiang Chen, Guangwei Gao, Jinshan Pan, Siyuan Yao, Shujun Fu, +1 more

Organizations: Shandong University · Central South University · Nanjing University of Science and Technology · Beijing University of Posts and Telecommunications · National University of Defense Technology

Abstract

Existing video deraining methods typically exploit neighboring frames through either explicit alignment or implicit spatiotemporal aggregation. Explicit alignment relies on accurate motion estimation, which can become unreliable under dense rain, while implicit aggregation avoids alignment but lacks explicit guidance on the directional and temporally coherent structure of rain. This leaves a gap between reliable temporal aggregation and explicit modeling of rain motion. To address these limitations, we propose FluidRain, a lightweight video derainer that uses divergence-free rain flow to guide Loop-in-Loop attention across scales and neighboring frames. Motivated by fluid mechanics, we model rain motion as a divergence-free image-space flow and use it to organize multi-scale and temporal aggregation. Specifically, FluidRain first estimates a rain-flow field for each frame and projects it onto the divergence-free subspace. The resulting flow steers window attention along rain streaks, enabling neighboring frames to be aggregated without explicit alignment. Since rain-flow structure is preserved across scales and nearby frames, Loop-in-Loop reuses the same attention operator across both dimensions, resulting in a three-frame model with only 0.80M parameters. Experiments on four benchmarks show that FluidRain remains competitive with substantially larger restoration models. We further examine how temporal evidence scales with different input views. To evaluate whether the model remains reliable when rain motion changes across frames, we introduce RainSyn-Gust, which injects controlled changes in rain-streak direction into existing benchmarks. We also develop a physics-based no-reference metric that evaluates real-rain removal without requiring clean targets.

Figures & tables

Figure 1: Physical motivation and Loop-in-Loop design of FluidRain. (a) Rain flow guides anisotropic attention. (b) Temporal persistence motivates the inner loop across frames. (c) Scale consistency motivates the outer loop across resolutions.

				NTURain			RainSynLight			RainSynComplex			RainSynAll100
Method	Venue	#P (M)	Time (ms) $\downarrow$	PSNR $\uparrow$	SSIM $\uparrow$	LPIPS $\downarrow$	PSNR $\uparrow$	SSIM $\uparrow$	LPIPS $\downarrow$	PSNR $\uparrow$	SSIM $\uparrow$	LPIPS $\downarrow$	PSNR $\uparrow$	SSIM $\uparrow$	LPIPS $\downarrow$
SwinIR-light	ICCVW’21	0.91	803	32.79	0.952	0.0572	33.25	0.956	0.0667	26.61	0.848	0.2232	32.49	0.924	0.1150
Restormer	CVPR’22	26.11	232	35.09	0.962	0.0355	37.19	0.978	0.0241	30.07	0.897	0.1302	38.06	0.965	0.0417
NAFNet	ECCV’22	29.16	20	33.88	0.956	0.0432	35.60	0.967	0.0443	28.04	0.854	0.1980	35.30	0.934	0.1007
IDT	TPAMI’23	16.42	249	33.23	0.952	0.0586	35.81	0.969	0.0404	27.97	0.860	0.1894	35.44	0.946	0.0733
DRSformer	CVPR’23	33.66	468	34.07	0.958	0.0474	37.46	0.977	0.0244	29.68	0.892	0.1356	37.41	0.958	0.0549

Table 1: Quantitative comparison on video deraining benchmarks. Best and second-best results are highlighted in bold and underline.

Figure 2: Qualitative comparison on video deraining benchmarks.

	Baseline	Repeat	Complementary views			Temporal range		Weight Sharing	Full model
Setting	Centre Only	Repeat Centre	Shifted Centre	One neighbour	Neighbour + shift	Two neighbours	Five frames	Untied scales	FluidRain
Passes	$t$	$t,t$	$t,t$	$t,t{+}1$	$t,t{+}1$	$t,t{\pm}1$	$t,t{\pm}1,t{\pm}2$	$t,t{\pm}1$	$t,t{\pm}1$
#P (M)	0.79	0.79	0.79	0.79	0.79	0.79	0.79	2.17	0.80
NTURain	32.57	32.73	33.06	33.54	33.78	34.55	34.25	33.81	35.73
RainSynLight	33.29	33.17	33.79	35.65	35.66	36.85	36.71	36.11	37.52
RainSynComplex	26.56	26.61	27.47	27.85	27.96	28.78	29.04	28.55	30.18

Table 2: Analysis of the Loop-in-Loop design.

Variant	#P (M)	NTURain	RainSynLight	RainSynComplex	RainSynAll100
w/o rain-flow pathway	0.79	34.55	36.85	28.78	34.72
w/o projection	0.80	35.01	37.06	29.73	35.14
w/o rain-flow bias	0.80	35.07	36.96	29.61	35.08
w/o differential term	0.80	35.29	37.23	30.07	35.53
FluidRain (full)	0.80	35.73	37.52	30.18	35.64

Table 3: Ablation of the rain-flow pathway.

Figure 3: Intervention on the rain-flow attention bias. Removing the bias weakens the flow-aligned attention, while rotating the flow redirects it.

Figure 4: Robustness to changing rain motion on RainSyn-Gust. The bottom strip shows the controlled $30^{\circ}$ perturbation and the corresponding post-turn restoration errors.

Figure 5: Evaluation of $F_{\mathrm{phys}}$ . (a) Correlation with PSNR on NTURain. (b) Precision–recall on real rain. (c) Restored crops and residuals.

Figure 6: Downstream perception under rain.

Explore similar work

Jul 15, 2026cs.CV

RainDancer: RGB-Event Video Deraining with Rain-Oriented Spiking Dynamics

Video deraining aims to recover clean visual content from rainy videos for reliable perception under adverse weather. Existing methods mainly rely on RGB sequences and temporal redundancy, but RGB-only restoration remains ambiguous in dynamic rainy scenes, where rain streaks, textures, boundaries, motion, and occlusions may share similar visual patterns. Event cameras provide complementary motion-sensitive cues with high temporal resolution, but event streams also contain sensor noise and background-triggered responses, so direct RGB-Event fusion may introduce cross-modal interference. To address this issue, we propose RainDancer, a progressive RGB-Event video deraining framework based on a decompose-before-interact paradigm. The core idea is to separate rain and background components within each modality before cross-modal interaction. In the RGB branch, frame features are progressively decomposed into rain and background representations. In the event branch, a rain-oriented spiking neural network module captures sparse and bursty event dynamics associated with rain motion. Component-level fusion is then performed between semantically aligned representations for structure preservation and rain suppression. We further introduce event-domain supervision to regularize sparse event reconstruction, structural consistency, and gradient orientation. Experiments on synthetic and real RGB-Event video deraining datasets demonstrate superior quantitative performance, visual quality, and downstream perception robustness. Code is available at https://github.com/AE86-plus/RainDancer.

Kui Jiang, Runzhe Li, Zhaocheng Yu +3

School of Computer Science and Technology, Harbin Institute of Technology, Harbin, China · School of Computer Science and Technology, Harbin University of Science and Technology, Harbin, China

Rgbt Video Restoration Fine-Grained Video Understanding Process-Level Supervision

Ask PDF

Aug 3, 2026cs.CV

DerainSplat: Feed-Forward Clean 3D Gaussian Splatting from Sparse Rainy Views

Although image deraining has advanced substantially, existing methods mainly focus on 2D image restoration. As spatial intelligence applications such as embodied AI and autonomous driving continue to emerge, reconstructing clean 3D scenes from sparse rainy views in a feed-forward manner becomes increasingly important. Existing feed-forward 3D Gaussian Splatting (3DGS) methods often assume clean inputs and collapse under rainy conditions. To this end, we present \textbf{\textit{DerainSplat}}, a feed-forward framework that reconstructs clean 3D scenes from only a few rainy views. To support this task, we build a large-scale multi-view derain dataset through a four-stage synthesis pipeline that sequentially models overcast illumination, depth-dependent haze, rain streaks, and lens raindrops, producing privileged weather factors. We introduce a weather net that predicts the weather factors from rainy context and yields two support maps. Scene support modulates cross-view cost-volume matching, while radiance support drives depth-aligned appearance fusion to fill corrupted pixels. The derived geometry evidence further attenuates Gaussian opacity to reduce spurious structures. A rainy cycle consistency re-renders clean views using the predicted factors and aligns them with rainy inputs. Extensive experiments show that \textbf{\textit{DerainSplat}} outperforms existing methods on various datasets, including RealEstate10K, ACID, Mip-NeRF360, and real-world rainy scenes, with strong cross-dataset generalization.

Fuzhen Jiang, Changyue Shi, Chuxiao Yang +3

Hangzhou Dianzi University · School of AI for Science, Peking University

Feed-Forward 3D Gaussian Splatting Feed-Forward 3D Gaussian 3D Generation Omnidirectional Images Multi-View

Ask PDF

Apr 21, 2026cs.CV

IR-Flow: Bridging Discriminative and Generative Image Restoration via Rectified Flow

In image restoration, single-step discriminative mappings often lack fine details via expectation learning, whereas generative paradigms suffer from inefficient multi-step sampling and noise-residual coupling. To address this dilemma, we propose IR-Flow, a novel image restoration method based on Rectified Flow that serves as a unified framework bridging the gap between discriminative and generative paradigms. Specifically, we first construct multilevel data distribution flows, which expand the ability of models to learn from and adapt to various levels of degradation. Subsequently, cumulative velocity fields are proposed to learn transport trajectories across varying degradation levels, guiding intermediate states toward the clean target, while a multi-step consistency constraint is presented to enforce trajectory coherence and boost few-step restoration performance. We show that directly establishing a linear transport flow between degraded and clean image domains not only enables fast inference but also improves adaptability to out-of-distribution degradations. Extensive evaluations on deraining, denoising and raindrop removal tasks demonstrate that IR-Flow achieves competitive quantitative results with only a few sampling steps, offering an efficient and flexible framework that maintains an excellent distortion-perception balance. Our code is available at https://github.com/fanzh03/IR-Flow.

Zihao Fan, Xin Lu, Jie Xiao +3

MoE Key Laboratory of Brain-inspired Intelligent Perception and Cognition, School of Information Science and Technology, University of Science and Technology of China

Image Restoration Rectified Flow Discriminative

Ask PDF