信息与控制2025，Vol.54Issue(2)：299-309,320,12.DOI:10.13976/j.cnki.xk.2024.4093
基于轻量模块化设计的水下图像增强方法
Underwater Image Enhancement Method Based on Lightweight Modular Design
摘要
Abstract
Underwater imaging technology faces challenges such as reduced contrast and color shifts caused by light absorption and scattering in marine environments,Which is particularly evident in image dependent applications in marine and underwater environments.To overcome these challen-ges,we introduce a framework called a lightweight modular underwater network(LMUW-Net).This framework is designed to improve the visual quality and color fidelity of underwater images without significantly increasing network complexity.This framework first enhances the structural features of underwater images through basic feature extraction.It then employs advanced three-channel colors(red,green,and blue)enhancement and global sparse feature enhancement to ef-fectively extract global image information,enhancing contrast and brightness in degraded underwa-ter images.LMUW-Net significantly reduces computational complexity and memory usage through network structure optimization and parameter-sharing mechanisms,enabling real-time processing while significantly improving visual quality and color fidelity.Extensive experimental results across multiple publicly available underwater image datasets have shown that LMUW-Net,with only 9 000 trainable parameters,improves image quality indicators such as peak signal-to-noise ratio by 5%and structural similarity index by 3%compared to existing methods.This highlights its significant advantages in enhancing underwater image visual quality and computational efficiency.Overall,LMUW-Net provides a robust solution for underwater image processing,especially suitable for ap-plications requiring real-time processing by improving image clarity.关键词
图像增强/深度学习/轻量化网络Key words
image enhancement/deep learning/lightweight network分类
信息技术与安全科学引用本文复制引用
牛奔,李亿平,周天薇,高宏伟..基于轻量模块化设计的水下图像增强方法[J].信息与控制,2025,54(2):299-309,320,12.基金项目
国家自然科学基金项目(72334004,62103286,71971143) （72334004,62103286,71971143）
广东省自然科学基金项目(2024A1515030278,2024A1515011712,2025A1515012829) （2024A1515030278,2024A1515011712,2025A1515012829）
广东省哲学社会科学十四五规划项目(GD22CGL35) （GD22CGL35）
广东省普通高校重点领域专项(2022ZDZX2054) （2022ZDZX2054）
广东省创新团队资助项目(2021WCXTD002) （2021WCXTD002）
深圳市自然科学基金面上项目(JCYJ20240813141612017) （JCYJ20240813141612017）
