船舶多聚焦图像的融合方法研究  

Research on the fusion method of multi-focus image of ship

作　　者：郝春云[1] HAO Chunyun(Nanchang Institute of Technology,Nanchang 330044,China)

机构地区：[1]南昌理工学院,江西南昌330044

出　　处：《舰船科学技术》2024年第15期181-184,共4页Ship Science and Technology

摘　　要：受到环境因素以及抖动等原因,船舶相机获取的图片存在模糊的现象,通过对多张图片进行融合可以有效解决这个问题。本文提出一种基于深度学习和多聚焦图像的融合方法,详细分析了多聚焦图像融合的过程,阐述了基于SIFT变换的特征提取原理,提出基于CNN的多聚焦图像融合训练阶段网络框架和图像融合网络框架,并进行实证研究,结果表明本文提出的多聚焦图像融合方法可以实现多张图片的有效融合,并在空间尺度上实现了坐标统一,输出图片具有较高的清晰度。Due to environmental factors and vibrations,images captured by shipboard cameras tend to be blurry.This issue can be effectively addressed by fusing multiple images.This paper proposes a deep learning-based and multi-focus image fusion method,thoroughly analyzes the process of multi-focus image fusion,elaborates on the feature extraction principle based on SIFT transformation,and presents the network framework for the training phase of CNN-based multi-focus image fusion and the image fusion network framework.Empirical research has been conducted,and the results demonstrate that the proposed multi-focus image fusion method can effectively integrate multiple images,achieve spatial scale coordination,and produce output images with high clarity.

关 键 词：SIFT CNN 多聚焦图像融合 船舶 

分 类 号：U661[交通运输工程—船舶及航道工程]