Underwater Image Restoration via Polymorphic Large Kernel CNNs

TL;DR

Underwater image restoration faces severe degradation from optical properties of water. The authors present UIR-PolyKernel, a lightweight pure-CNN approach that uses polymorphic large kernels and a Hybrid Domain Attention module to capture long-range and hidden features with high efficiency. Key components include Composite Shape Convolution (CSC), Large Kernel Attention (LKA), and Hybrid Domain Attention (HDA) with a frequency-domain attention pathway, trained via a composite loss that combines pixel, structural, and underwater quality terms. The method achieves state-of-the-art results on UIEB, EUVP, and LSUI with favorable computational cost, demonstrating that carefully designed CNNs can rival more complex architectures for underwater restoration and enable practical, real-time deployment.

Abstract

Underwater Image Restoration (UIR) remains a challenging task in computer vision due to the complex degradation of images in underwater environments. While recent approaches have leveraged various deep learning techniques, including Transformers and complex, parameter-heavy models to achieve significant improvements in restoration effects, we demonstrate that pure CNN architectures with lightweight parameters can achieve comparable results. In this paper, we introduce UIR-PolyKernel, a novel method for underwater image restoration that leverages Polymorphic Large Kernel CNNs. Our approach uniquely combines large kernel convolutions of diverse sizes and shapes to effectively capture long-range dependencies within underwater imagery. Additionally, we introduce a Hybrid Domain Attention module that integrates frequency and spatial domain attention mechanisms to enhance feature importance. By leveraging the frequency domain, we can capture hidden features that may not be perceptible to humans but are crucial for identifying patterns in both underwater and on-air images. This approach enhances the generalization and robustness of our UIR model. Extensive experiments on benchmark datasets demonstrate that UIR-PolyKernel achieves state-of-the-art performance in underwater image restoration tasks, both quantitatively and qualitatively. Our results show that well-designed pure CNN architectures can effectively compete with more complex models, offering a balance between performance and computational efficiency. This work provides new insights into the potential of CNN-based approaches for challenging image restoration tasks in underwater environments. The code is available at \href{https://github.com/CXH-Research/UIR-PolyKernel}{https://github.com/CXH-Research/UIR-PolyKernel}.

Figures (2)

• Figure 1: The overall architecture of the proposed UIR-PolyKernel model. This model utilizes three resolution levels of feature maps, generated by two pairs of downsampling and upsampling operations. Each downsampling operation halves the spatial dimensions of the feature map while doubling the number of channels, with upsampling operations performing the inverse. The initial number of channels in the first level of feature maps is 36.
• Figure 2: Visual comparison of the proposed method with two leading traditional approaches and the three top-performing deep learning methods.