Multilevel approach for bright spot removal, fine detail restoration, and color correction in underwater imagery

The enhancement of underwater imagery has garnered considerable interest due to diverse applications across fields such as marine engineering and aquatic robotics. Artificial lights (ALs) enhance visibility and minimal color distortion in underwater conditions. However, AL introduces non-uniform illumination that leads to degradation of the overall image quality. Most existing underwater image enhancement algorithms tend to neglect this specific phenomenon. Their observations indicate that areas affected by AL frequently display tendencies toward artifacts and detail loss, adversely affecting the enhancement process’s overall efficacy. A underwater image enhancement model for illumination, reflection, and AB color channel correction (UIRAB) model is proposed to enhance underwater images affected by artificial lighting by balancing illumination and reflection artifacts in the L channel and correcting colors in the AB channels using the luminance and AB (LAB) color space. The presented UIRAB model integrates three distinct modules: illumination correction, reflection correction, and AB channel correction. These modules are proposed for light correction, fine details enhancement, and color restoration, respectively. Experiments are performed on four datasets: real underwater image dataset with AL, large-scale underwater image dataset, underwater image enhancement benchmark dataset, and enhancing underwater visual perception dataset. The experimental results demonstrate that the proposed model surpasses existing state-of-the-art approaches in enhancing synthetic and natural underwater images affected by AL.