A Comprehensive Review of Fusion-Based Underwater Image and Video Enhancement: Techniques, Challenges, and Emerging Applications

Underwater image and video enhancement has emerged as a critical research frontier in computer vision and marine technology, driven by optical degradation phenomena inherent to aquatic environments. Light absorption and scattering in water produce images that are characteristically hazy, low-contrast, and color-distorted, severely limiting the utility of underwater imaging for scientific, industrial, and exploratory purposes. This review critically examines the landscape of underwater image enhancement methodologies, with particular emphasis on single-image fusion-based strategies that operate without specialized hardware or a priori knowledge of scene geometry. We survey the physical underpinnings of underwater optical degradation, evaluate the comparative limitations of classical enhancement methods including white balancing, histogram equalization, and polarization-based restoration, and provide an in-depth analysis of multi-scale Laplacian pyramid fusion frameworks. Weight map design—encompassing Laplacian contrast, local contrast, saliency, and exposedness measures—is analyzed for its role in perceptually motivated enhancement. We also assess temporal coherence strategies using bilateral filtering for video applications. Recent developments in deep learning-based methods, including convolutional neural networks, generative adversarial networks, and vision transformers, are reviewed in the context of underwater enhancement. The review situates these methods within broader application domains such as feature-point matching, image segmentation, and single-image dehazing. Limitations of current approaches and promising research directions are discussed. This review aims to serve as a foundational reference for researchers and practitioners advancing the state of the art in underwater visual computing.