CLUIE: Clustering-Aware Recurrent Propagation with Local Structural Compensation for Underwater Image Enhancement

2026-07-23 · Source: Computer Vision and Pattern Recognition · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Computer Vision · Depth: Expert, quick

Summary

CLUIE, a Clustering-aware Recurrent Propagation framework with Local Structural Compensation, addresses challenges in underwater image enhancement. Traditional methods struggle with heterogeneous degradation patterns caused by wavelength-dependent light absorption, scattering, and backscattering, leading to color distortion, contrast degradation, and detail loss. While visual RWKV models efficiently handle long-range dependencies, their fixed scanning orders fail to adapt to non-uniform restoration demands. The proposed CRWKV framework reformulates RWKV's fixed recurrent propagation into a content-adaptive token trajectory. It introduces Clustering-aware Semantic Dynamic Reordering (CSDR) to group tokens by semantic similarity and derive dynamic traversal orders based on inter-cluster relations, enabling WKV state accumulation along semantically correlated regions. Additionally, Dark-response Modulated Local Propagation (DMLP) extracts local structural responses via depth-wise convolution, modulating their strength with a pseudo-dark response map to compensate local structural cues. Experiments on multiple benchmarks confirm CRWKV achieves state-of-the-art quantitative performance and superior visual quality.

Key takeaway

For Computer Vision Engineers developing underwater imaging solutions, you should recognize that fixed-order recurrent models are suboptimal for heterogeneous degradation patterns. Your projects can achieve superior visual quality by adopting content-adaptive propagation frameworks like CRWKV. Specifically, integrate semantic clustering for dynamic token reordering and incorporate local structural compensation mechanisms. This approach directly addresses region-specific degradation, offering a significant improvement over conventional methods and enhancing the reliability of your underwater vision systems.

Key insights

Content-adaptive recurrent propagation and local structural compensation enhance underwater images by addressing heterogeneous degradation.

Principles

• Degradations in underwater images are region-adaptive.
• Recurrent propagation benefits from content-adaptive paths.
• Local structural cues need compensation in dynamic reordering.

Method

CRWKV uses Clustering-aware Semantic Dynamic Reordering (CSDR) for content-adaptive token trajectories and Dark-response Modulated Local Propagation (DMLP) to compensate local structural cues before recurrent aggregation.

In practice

• Apply semantic clustering for adaptive image processing.
• Integrate local structural compensation in recurrent models.
• Consider dynamic reordering for non-uniform data.

Topics

• Underwater Image Enhancement
• Recurrent Neural Networks
• Computer Vision
• Image Degradation
• Clustering Algorithms
• Semantic Reordering

Best for: Research Scientist, AI Scientist, Computer Vision Engineer, Machine Learning Engineer

Related on AIssential

• Fusing Transferred Priors and Physics-based Decomposition for Underwater Image Enhancement — A novel UIE method fuses transferred priors and physics-based decomposition to enhance underwater images without paired labels.
• ReA-OVCD: Reliability-Aware Open-Vocabulary Change Detection via Semantic and Spatial Refinement — ReA-OVCD refines open-vocabulary change detection by collaboratively addressing semantic ambiguity and spatial inconsistency through a training-free, two-module approach.
• DART: A Degradation-Aware Recurrent Transformer for Archival Film Restoration — DART explicitly guides film restoration by predicting and propagating soft defect masks.
• A Probabilistic Framework for Improving Dense Object Detection in Underwater Image Data via Annealing-Based Data Augmentation — A pseudo-simulated annealing data augmentation method improves underwater object detection in crowded, natural scenes.
• Plug-and-Play image restoration with Stochastic deNOising REgularization — SNORE improves PnP image restoration by applying denoisers to appropriately noisy images, enhancing performance and theoretical guarantees.
• WAT3R: Feedforward Underwater 3D Reconstruction — WAT3R is a feed-forward framework that uses a neural adaptation module to improve underwater 3D reconstruction by accounting for imaging degradation.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.