结合Otsu阈值优化和多尺度融合的海上图像去雾算法

Dehazing algorithm of martime images combining Otsu algorithm threshold optimization and multi-scale fusion

扫码查看

原文链接

万方数据
维普

中文摘要：针对在处理海上图像时暗通道先验去雾算法对天空区域不适用以及复原图像易出现色偏的问题,提出一种结合天空区域分割与多尺度透射率融合的海上图像暗通道去雾算法.采用灰度和双边滤波预处理图像,基于大津算法(Otsu)阈值优化对预处理后的图像进行天空区域初分割,再进行基于梯度域导向滤波的天空区域细分割,并去除细分割图像中的阴影;基于天空区域分割改进大气光值估计方法,通过多尺度透射率融合获得初始透射率图,以及采用梯度域导向滤波细化透射率图;通过大气散射模型完成对海上图像的去雾处理.结果表明,与原始暗通道去雾算法相比,利用所提算法去雾后的图像在信息熵、结构相似度、峰值信噪比指标上分别提升了 3.5％、9.89％、27.42％.该算法能够有效避免原始算法产生的光晕和色偏现象,且具有更好的视觉效果,能还原真实的海上场景船舶图像.

外文摘要：Aiming at the problem that the dark channel prior dehazing algorithm is not applicable to the sky region and restored images are prone to color errors when processing martime images,a dark channel dehazing algorithm for maritime images combined with sky region segmentation and multi-scale transmittance fusion is proposed.The images are pre-processed by grayscale and bilateral fitering.The sky region in the pre-processed images is initially segmtioned based on the Otsu algorithm threshold optimization,the fine segmentation of sky region is performed based on gradient domain guided filtering,and the shadow in the fine-segmented images is removed;the estimation method of atmospheric light value is improved based on the sky region segmentation,the initial transmittance image is obtained by multi-scale transmittance fusion,and the gradient domain guided filtering is used to refine the transmittance image;the dehazing process of the maritime images is completed through the atmospheric scattering model.The results show that,compared with the original dark channel dehazing algorithm,the information entropy,structural similarity,and peak signal-to-noise ratio of the images which are dehazed by the proposed algorithm are improved by 3.5％,9.89％,and 27.42％,respectively.The algorithm can avoid the halo and color errors resulting from the original algorithm effectively,has a better visual effect,and can restore the real ship image of the sea scene.

外文关键词：

maritime imageimage dehazingsky segmentationmulti-scale fusiongradient domain guided filteringdark channel prior algorithm

作者：

迟明善、刘鹏杰、张强、弭永发、宁心怡

展开 >

作者单位：

山东交通学院航运学院,山东威海 264209

关键词：

海上图像图像去雾天空分割多尺度融合梯度域导向滤波暗通道先验算法

基金：

山东省自然科学基金山东省大学生创新创业训练计划山东交通学院博士科研启动基金2023年山东交通学院研究生科技创新项目

项目编号：

ZR2022ME087S202311510240BS20240092023YK098

出版年：

2024

DOI：

10.13340/j.jsmu.202306120132

上海海事大学学报

上海海事大学

上海海事大学学报

CSTPCD北大核心

影响因子：0.578

ISSN：1672-9498

年,卷(期)：2024.45(3)

参考文献量13