Advanced Underwater Image Enhancement and Object Detection using YOLOv2

Underwater image processing presents significant challenges due to scattering, low contrast, and color distortion. This paper proposes an integrated approach combining low-frequency image extraction, dehazing using Dark Channel Prior, white balancing, and high-frequency amplification to enhance underwater images. Additionally, discrete wavelet transform (DWT) and Laplacian decomposition are applied to extract high-frequency details. A contrast adjustment step and non-local means filtering further refine the image quality. Finally, deep learning-based object detection using YOLOv2 is applied to the enhanced images, improving object recognition in underwater environments. Experimental results demonstrate improved perceptual quality and object detection accuracy. Key performance metrics such as PSNR, entropy, and NIQE validate the proposed approach.