Aqua-Vision: The Future of Underwater Exploration and Innovation

Abstract

Underwater photos are afflicted with color distortion, low contrast, haze, and noise, which compromise visual quality and affect tasks such as object detection. Current approaches typically handle one form of degradation and the resulting system cannot be as effective. We introduce a single-system, unified frame work that unifies color correction, dehazing, denoising, and object detection. The strategy blends traditional methods (histogram equalization, Dark Channel Prior) with deep learning approaches (CNN-based denoising, YOLO detection) to recover natural colors, improve details, eliminate noise, and precisely detect objects. Performance assessment with metrics like PSNR, SSIM, UISM, BRISQUE, NIQE, UICM, UCIQE, CIEDE2000, and mAP exhibits marked improvement in both computational performance and visual quality, making the system appropriate for real-time underwater applications such as marine monitoring and autonomous navigation.