基于图像熵线性加权的水下图像增强算法  

An Underwater Image Enhancement Algorithm Based on Image Entropy Linear Weighting

作　　者：王亚强 吴明晖 耿方琪 冯业宁 周围 WANG Yaqiang;WU Minghui;GENG Fangqi;FENG Yening;ZHOU Wei(School of Mechanical and Automotive Engineering,Shanghai University of Engineering and Science,Shanghai 201600,China)

机构地区：[1]上海工程技术大学机械与汽车工程学院,上海201600

出　　处：《重庆工商大学学报（自然科学版）》2024年第4期69-76,共8页Journal of Chongqing Technology and Business University:Natural Science Edition

基　　金：上海市自然科学基金项目(21ZR1425900)。

摘　　要：目的针对工业环境下水下图像受到水中悬浮物影响,从而导致图像的清晰度过低以及对比度过高等问题,提出一种基于图像熵线性加权的水下图像增强算法。方法该算法基于图像熵理论对白平衡算法、直方图均衡算法和暗通道先验算法进行线性加权,继而通过实验环境确定调节系数输出高质量图像。在深度为1 m、1.5 m和2 m的不同水下环境拍摄图像,对获得的水下图像使用上述三种算法和该算法作对比处理,处理结果通过PSNR和UIQM作为评价指标进行评判。结果实验结果表明:使用PSNR指标评判该算法,相较于其他三种算法,水深1 m的水下图像质量提高了22.81%,水深1.5 m的水下图像质量提高了46.67%,水深2 m的水下图像质量提高了38.94%,图像质量综合平均提高了36.14%;使用UIQM指标评判该算法,相较于其他三种算法,水深1 m的图像质量提高了1.02%,水深1.5 m的水下图像质量提高了0.73%,水深2 m的水下图像质量提高了1.82%,图像质量综合平均提高了1.19%。结论由此可以证明该算法相对于其他传统算法对图像清晰度有着显著提升,并且能够适应不同深度的水下环境,为工业环境下水下图像增强提供了一种新的解决思路。Objective Aiming at the problem that the underwater images in the industrial environment are affected by suspended solids in water,resulting in low image clarity and high contrast,an underwater image enhancement algorithm based on linear weighting of image entropy was proposed.Methods The algorithm linearly weighted the white balance algorithm,histogram equalization algorithm,and dark channel prior algorithm based on the image entropy theory,and then determined the adjustment coefficient to output high-quality images through the experimental environment.Through taking images in different underwater environments at depths of 1 m,1.5 m,and 2 m,the obtained underwater images were compared and processed by using the above three algorithms and this algorithm.The processing results were evaluated by PSNR and UIQM as evaluation indicators.Results The experimental results showed that the proposed algorithm,judged by the PSNR index,improved the underwater image quality by 22.81%for the water depth of 1m,46.67%for the water depth of 1.5 m,38.94%for the water depth of 2 m,and improved the combined image quality by 36.14%on average,compared with the other three algorithms.The proposed algorithm,judged by the UIQM index,improved the image quality by 1.02%for the water depth of 1m,0.73%for the water depth of 1.5 m,1.82%for the water depth of 2 m,and improved the combined image quality by 1.19%on average,compared with the other three algorithms.Conclusion It can be proved that the proposed algorithm has significantly improved the image definition compared with other traditional algorithms,and this algorithm can adapt to different depths of underwater environments,providing a new solution for underwater image enhancement in industrial environments.

关 键 词：水下图像增强 线性加权 图像熵 图像融合 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]