Vision mamba improves underwater images by isolating update states
State-Isolated Vision Mamba via Update-Tendency Routing for Underwater Image Enhancement
Computer Vision and Pattern Recognition
Summary
Underwater photos often look blurry or washed out, and enhancing them is tricky because different parts of the image need different adjustments. The authors found that if all parts of an underwater image try to update the same shared information in a model, they interfere with each other, making enhancement less effective. They designed a new method called IsoMamba that groups parts of the image that need similar updates and processes them separately to avoid this interference. This approach leads to clearer underwater images and better overall enhancement performance.
What this means in practice
- •For underwater photographers: Enhance underwater photos automatically with improved image clarity by handling different regions with isolated processing streams.
- •For marine robotics engineers: Improve underwater robot vision systems by applying state-isolated recurrent models to enhance real-time visual sensing in challenging aquatic environments.
Authors
Yan Luo, Laibin Chang, Zefan Feng, Yi'ang Chen, Kui Jiang, Zhengguo Li
Abstract
Vision Mamba has recently emerged as an efficient alternative to convolutional and Transformer-based architectures by modeling long-range spatial dependencies through recurrent state evolution. Existing studies primarily improve token serialization, scanning strategies, or expert routing, yet largely overlook a state level question question: how should recurrent states be organized for spatially heterogeneous visual restoration? In this paper, we show that forcing tokens with heterogeneous update behaviors to recurrently modify a shared hidden state leads to persistent cross-content interference, which we term \textbf{State Contention}. We further argue that effective recurrent modeling should organize state evolution according to token compatibility rather than token ordering. Based on this observation, we introduce \textbf{State Isolation}, a new recurrent state organization principle for Vision Mamba,tokens with compatible state-update tendencies evolve within independent recurrent state streams. To instantiate this principle, we propose \textbf{IsoMamba}, a $Δ$-guided architecture that constructs model-intrinsic update representations, discovers update-compatible token groups through label-free clustering, and routes them to isolated Mamba experts following a \textbf{cluster--route--isolate} paradigm. In addition, we develop a dynamic channel-wise reweighting objective to compensate for wavelength-dependent reconstruction difficulty in underwater images. Extensive experiments demonstrate that IsoMamba consistently achieves state-of-the-art performance on multiple underwater image enhancement benchmarks while providing cleaner hidden state responses, validating the effectiveness of state isolated recurrent modeling.