Underwater Image Enhancement Based on Adaptive Color Correction and Multi-scale Fusion

Underwater imaging is characterized by significant color distortion and low contrast, primarily due to the complex interactions of light absorption and scattering in aquatic environments. This work presents a novel enhancement framework designed to address these challenges through adaptive optimization and advanced fusion techniques. The method operates independently of specialized hardware and predefined environmental conditions. The enhancement process begins with adaptive color correction to mitigate color cast, followed by a transformation to the HSV color space for isolated value (V) channel processing. Principal Component Analysis (PCA) integrates the processed components, yielding a globally enhanced image. Simultaneously, image decomposition refines fine-scale details, and an enhanced Retinex algorithm is applied to accentuate edge structures. The Non-Subsampled Shearlet Transform (NSST) fuses the globally enhanced, detail-enhanced, and edge-enhanced images into a unified output, followed by further optimization to produce the final enhanced image. Extensive experiments confirm superior performance across various metrics, including PCQI, UCIQE, UIQM, and IE, surpassing existing state-of-the-art methods. Feature preservation capabilities are validated using the Speeded-Up Robust Features (SURF) algorithm, highlighting the method’s potential in underwater exploration, photography, and robotic vision.