MTNet: Multi-Task Underwater Image Enhancement Method Based on Retinex

• DOI: 10.23967/j.rimni.2024.10.60509
• Publication Date: 2025-05-13T14:44:10Z
• Authors: Y. Song; X. Deng; H. Shao; F. Wang
• Keywords: Color Constancy

Underwater images play a critical role in underwater exploration and related tasks. However, due to light attenuation other factors, often suffer from color distortion low contrast, which some extent limit the efficiency safety of exploration. To meticulously address these issues enhance accuracy reliability exploration, this paper proposes multi-task image enhancement method based on Retinex theory. This divides task into several sub-tasks, including decomposition, correction, detail reconstruction, illumination adjustment. Specialized sub-networks— DecomNet, DecolorNet, DelightNet—are designed specifically problems, thereby alleviating distortion, enhancing details, improving contrast. Experiments conducted publicly datasets indicate that quality is significantly improved after with proposed method, compared representative processing techniques. For example, real-world dataset Image Enhancement Benchmark, MSE, Structural Similarity Index Measure, Peak signal-to-noise ratio scores achieved were 453.480, 0.901, 25.145, respectively. study holds significant implications for potential applications fields marine research archaeology.OPEN ACCESS Received: 03/11/2024 Accepted: 27/12/2024 Published: 20/04/2025