Enhancement of Underwater Images using DCP and CycleGAN

The enhancement of underwater images has seen significant advancements with the introduction of various methods and datasets. This paper presents a novel approach combining the Dark Channel Prior (DCP) and Cycle-Consistent Generative Adversarial Network (CycleGAN) to address common issues in underwater imagery, such as haze and color distortion. By first applying DCP to reduce haze and then utilizing CycleGAN for color correction and enhancement, our model produces superior results compared to traditional methods. Specifically, our approach proves highly effective on haze images, significantly improving their clarity and color fidelity. We evaluated its performance using the Underwater Image Quality Measure (UIQM), achieving a total score of 3.3137, demonstrating substantial improvements in color accuracy, contrast, and overall clarity, and surpassing existing models. The model’s effectiveness is validated on diverse datasets, showcasing its potential to enhance underwater images across various conditions.