Search papers, labs, and topics across Lattice.
This paper introduces a real-time underwater image enhancement framework that prioritizes color accuracy for improved visual reliability in underwater environments. The framework incorporates three key modules: Adaptive Weighted Channel Compensation for dynamic color recovery, Multi-branch Re-parameterized Dilated Convolution for efficient large receptive field representation, and Statistical Global Color Adjustment for overall color optimization. Experiments on eight datasets demonstrate state-of-the-art performance across seven evaluation metrics, achieving 409 FPS with only 3,880 parameters and a 29.7% improvement in UCIQE score.
Achieve real-time (409 FPS) underwater image enhancement with a tiny (3,880 parameter) model that significantly improves color accuracy, enabling deployment on resource-constrained underwater platforms.
Underwater image enhancement plays a crucial role in providing reliable visual information for underwater platforms, since strong absorption and scattering in water-related environments generally lead to image quality degradation. Existing high-performance methods often rely on complex architectures, which hinder deployment on underwater devices. Lightweight methods often sacrifice quality for speed and struggle to handle severely degraded underwater images. To address this limitation, we present a real-time underwater image enhancement framework with accurate color restoration. First, an Adaptive Weighted Channel Compensation module is introduced to achieve dynamic color recovery of the red and blue channels using the green channel as a reference anchor. Second, we design a Multi-branch Re-parameterized Dilated Convolution that employs multi-branch fusion during training and structural re-parameterization during inference, enabling large receptive field representation with low computational overhead. Finally, a Statistical Global Color Adjustment module is employed to optimize overall color performance based on statistical priors. Extensive experiments on eight datasets demonstrate that the proposed method achieves state-of-the-art performance across seven evaluation metrics. The model contains only 3,880 inference parameters and achieves an inference speed of 409 FPS. Our method improves the UCIQE score by 29.7% under diverse environmental conditions, and the deployment on ROV platforms and performance gains in downstream tasks further validate its superiority for real-time underwater missions.
