| 引用本文: | 覃贺权,宋磊,王成罡,于文彬.通道修正均衡化的水下图像增强算法[J].控制理论与应用,2022,39(11):2047~2056.[点击复制] |
| QIN He-quan,SONG Lei,WAMG Cheng-gang,YU Wen-bin.Underwater image enhancement algorithm based on channel correction equalization[J].Control Theory and Technology,2022,39(11):2047~2056.[点击复制] |
|
| 通道修正均衡化的水下图像增强算法 |
| Underwater image enhancement algorithm based on channel correction equalization |
| 摘要点击 782 全文点击 258 投稿时间：2021-11-29 修订日期：2022-04-27 |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| DOI编号 10.7641/CTA.2022.11165 |
| 2022,39(11):2047-2056 |
| 中文关键词 通道修正 限制对比度自适应直方图均衡化 暗通道先验 水下图像增强 |
| 英文关键词 channel correction contrast limited adaptive histogram equalization dark channel prior underwater image enhancement |
| 基金项目 国家自然科学基金项目(61773264, 61922058, 61803261, 61801295), 上海交通大学深蓝计划项目(SL2020ZD206, SL2020MS010, SL2020MS015) 资助. |
|
| 中文摘要 |
| 水中介质和微粒的影响导致光波传播衰减和散射, 在成像过程中水下图像会出现模糊和色偏等情况, 这些
水下成像退化的情况给水下的目标识别、目标跟踪、特征提取等应用带来困难. 针对以上问题, 本文提出了一种基
于通道修正均衡化的暗通道先验(CCD)水下图像增强算法: 首先是对色偏的水下图像进行通道修正均衡化, 利用直
方图强度中心做一个映射, 并将映射的三通道信息融合到限制对比度自适应直方图均衡化中, 改善了图像色偏和对
比度不足的情况; 其次是通过暗通道先验算法进行去模糊, 通过水下增强图像数据集的实验表明, CCD比现有算法
更有效地应对了水下图像成像退化问题, 取得了更好的图像质量指标; 此外, 在特征检测预处理步骤中, 本文方法能
够将检测特征点数提高约1.88倍. |
| 英文摘要 |
| Due to the influence of light scattering and attenuation in the underwater environment, images suffer from
color deviation and blur, making underwater target recognition, target tracking, feature extraction and other applications
difficult. We propose a channel correction equalization based on the dark channel prior underwater image enhancement
algorithm (CCD) to deal with the aforementioned problems. Firstly, the channel correction equalization is carried out for
the underwater image with color deviation. A mapping is made using the histogram intensity center subsequently, and the
mapped three channel information is fused into the limited contrast adaptive histogram equalization to improve the color
deviation and insufficient contrast. Secondly, a dark channel prior algorithm is used to deblur. Experimental results on
processing underwater image enhancement datasets demonstrate that the CCD outperforms current algorithms. In addition,
in the preprocessing step of feature extraction, the number of detected feature points can increase about 1.88 times. |
|
|
|
|
|