An Attention-Driven Multi-Scale Ensemble Framework for Underwater Image Enhancement

Abstract

Underwater image enhancement plays a critical role in improving image quality for applications such as marine exploration, underwater robotics, environmental monitoring, surveillance, and computer vision. However, underwater images often suffer from severe degradation caused by light absorption, scattering, low contrast, color distortion, and reduced visibility. To address these challenges, this study proposes an advanced underwater image enhancement framework that combines hybrid preprocessing techniques with deep learning-based enhancement strategies. Initially, Underwater White Balance (UWB) and Variational Contrast and Saturation Enhancement (VCSE) are applied to restore color balance and improve image contrast. The enhanced images are further processed using ensemble attention-driven convolutional networks, multi-scale feature fusion, and adaptive contrast-saturation optimization. Experimental evaluation using standard image quality metrics demonstrates significant improvement in color restoration, structural detail preservation, visibility, and overall enhancement performance across diverse underwater environments.