Improving Underwater Image Quality Through Real-ESRGAN With Whale Optimization Algorithm

Abstract

Unique optical properties of underwater environments, like low resolution, blurriness, and color distortion, are common challenges for underwater imaging. Consequently, the imaging equipment suffers from water turbidity, light attenuation, and scattering in aquatic environments, despite the improvement in hardware, resulting in lesser-quality, distorted, and poorly contrasted color images. An innovative approach to enhance underwater images by integrating Real-ESRGAN (Real-Enhanced Super-Resolution Generative Adversarial Network) with a Whale Optimization Algorithm (WOA) is studied in this research to address these issues. To fine-tune the model parameters and improve the overall image enhancement process, Real-ESRGAN, known for its superior performance in quality image resolution enhancement, is combined with WOA, a nature-inspired optimization algorithm. Extensive experiments on the LSUI dataset are conducted to evaluate the efficacy of this approach. The efficacy of the suggested approach is assessed comprehensively, combining qualitative visual analysis with quantitative metrics. The proposed method demonstrates strong quantitative performance, achieving a PSNR of 35.48, SSIM of 0.82, UIQM of 4.60, RMSE of 0.25, and entropy of 5.50. The outcomes indicate notable upgradation in image clarity, detail, and color accuracy compared to existing enhancement techniques. This research contributes to underwater imaging by offering an innovative solution that enhances the quality of underwater visuals.