CLEAR: An Efficient Low-Illumination Enhancement Method for Improved Visibility in Underwater Images

Enhancing underwater imagery is critical for marine research, environmental monitoring, and autonomous navigation, yet image acquisition in such environments is inherently challenging due to low illumination caused by light absorption, scattering from suspended particles, and microscopic organisms. These factors reduce visibility, distort colors, and obscure fine details, limiting both human interpretation and automated analysis. The Contrast and Light Enhancement for Aquatic Restoration (CLEAR) method addresses these issues by integrating multiple modules: the Low-Illumination Network (LINet) restores lost illumination using a synthetic dataset; the Edge Enhancement by Differencing (EED) module preserves edge structures; and Exponential Contrast Stretching (ECS) enhances sharpness and contrast. Additionally, the Low-Illumination Underwater Scenes (LIUS) dataset is introduced, generated by artificially adjusting images captured under normal lighting. Extensive experiments on an unpaired real-world dataset, Non-Uniform Illumination Dataset (NUID), show that CLEAR surpasses state-of-the-art methods in both visual quality and quantitative metrics. CLEAR also enhances the performance of downstream tasks, reinforcing its practical value and applicability in real-world scenarios. The code of this method is available at https://github.com/xingyumex/CLEAR.