Optimizing Underwater Image Enhancement using AquaFusion PH -Net

The multifaceted optical features of water, such as light attenuation and backscatter, present considerable obstacles for underwater image enhancement. Traditional techniques, such as histogram equalization and contrast stretching, often fail in underwater scenes because of the peculiar light absorption and scattering properties. The study, introducing a novel method, AquaFusion Patchwise Hybridization-Net (AquaFusion PH-Net), which implements a patchwise approach that allows for precise enhancement of local image details, ensuring optimized and adaptable improvements in underwater image quality. AquaFusion PH-Net significantly enhances underwater image processing by optimizing light attenuation to reduce backscatter effectively. This model employs mish activation for improved performance and utilizes leaky ReLU for advanced colour restoration. Furthermore, it incorporates non-local means for additional denoising, ensuring superior image quality across diverse underwater conditions. Experiments using more than 3,000 unpaired images demonstrate that the proposed model consistently outperforms existing enhancement methods, with metrics such as Structural Similarity Index (SSIM), Absolute Mean Brightness Error (AMBE), and Peak Signal-to-Noise Ratio (PSNR) indicating significant improvements in clarity, colour restoration, and artifact reduction.