Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PRISM: Progressive Rain removal with Integrated State-space Modeling

Pengze Xue, Shanwen Wang, Fei Zhou, Yan Cui, Xin Sun

TL;DR

PRISM addresses the deraining problem by introducing a progressive three-stage framework that integrates frequency-aware and global-context modeling. It combines Coarse Extraction Network (CENet) with Hybrid Attention UNet (HA-UNet), Frequency Fusion via SFNet, and a Refinement stage (RNet) that leverages an original-resolution subnetwork. The Hybrid Domain Mamba (HDMamba) fuses semantic reordering in the spatial domain with wavelet-domain processing, using adaptive gating to balance contributions across domains. A multi-component loss guides reconstruction and frequency fidelity across stages. Experiments on mixed synthetic datasets demonstrate competitive PSNR/SSIM and robust visual quality, validating PRISM's effectiveness across diverse rain patterns.

Abstract

Image deraining is an essential vision technique that removes rain streaks and water droplets, enhancing clarity for critical vision tasks like autonomous driving. However, current single-scale models struggle with fine-grained recovery and global consistency. To address this challenge, we propose Progressive Rain removal with Integrated State-space Modeling (PRISM), a progressive three-stage framework: Coarse Extraction Network (CENet), Frequency Fusion Network (SFNet), and Refine Network (RNet). Specifically, CENet and SFNet utilize a novel Hybrid Attention UNet (HA-UNet) for multi-scale feature aggregation by combining channel attention with windowed spatial transformers. Moreover, we propose Hybrid Domain Mamba (HDMamba) for SFNet to jointly model spatial semantics and wavelet domain characteristics. Finally, RNet recovers the fine-grained structures via an original-resolution subnetwork. Our model learns high-frequency rain characteristics while preserving structural details and maintaining global context, leading to improved image quality. Our method achieves competitive results on multiple datasets against recent deraining methods.

PRISM: Progressive Rain removal with Integrated State-space Modeling

TL;DR

PRISM addresses the deraining problem by introducing a progressive three-stage framework that integrates frequency-aware and global-context modeling. It combines Coarse Extraction Network (CENet) with Hybrid Attention UNet (HA-UNet), Frequency Fusion via SFNet, and a Refinement stage (RNet) that leverages an original-resolution subnetwork. The Hybrid Domain Mamba (HDMamba) fuses semantic reordering in the spatial domain with wavelet-domain processing, using adaptive gating to balance contributions across domains. A multi-component loss guides reconstruction and frequency fidelity across stages. Experiments on mixed synthetic datasets demonstrate competitive PSNR/SSIM and robust visual quality, validating PRISM's effectiveness across diverse rain patterns.

Abstract

Image deraining is an essential vision technique that removes rain streaks and water droplets, enhancing clarity for critical vision tasks like autonomous driving. However, current single-scale models struggle with fine-grained recovery and global consistency. To address this challenge, we propose Progressive Rain removal with Integrated State-space Modeling (PRISM), a progressive three-stage framework: Coarse Extraction Network (CENet), Frequency Fusion Network (SFNet), and Refine Network (RNet). Specifically, CENet and SFNet utilize a novel Hybrid Attention UNet (HA-UNet) for multi-scale feature aggregation by combining channel attention with windowed spatial transformers. Moreover, we propose Hybrid Domain Mamba (HDMamba) for SFNet to jointly model spatial semantics and wavelet domain characteristics. Finally, RNet recovers the fine-grained structures via an original-resolution subnetwork. Our model learns high-frequency rain characteristics while preserving structural details and maintaining global context, leading to improved image quality. Our method achieves competitive results on multiple datasets against recent deraining methods.

Paper Structure

This paper contains 11 sections, 19 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Hybrid Attention UNet
  4. Hybrid Domain Mamba
  5. Loss Function
  6. Experiments
  7. Datasets
  8. Evaluation Metrics
  9. Experimental Results
  10. Ablation Studies
  11. Conclusion

Figures (4)

  • Figure 1: HDMamba provides stronger global modeling and incorporates frequency-domain awareness.
  • Figure 2: The overall architecture of our proposed PRISM with CENet, SFNet, and RNet to achieve the rain-free image.
  • Figure 3: HA‑UNet architecture. A UNet encoder-decoder with hybrid attention and skip connections for feature fusion.
  • Figure 4: Deraining results on the Rain100L dataset yang2017deep. Close-up views highlight deraining details.