摘要
目的:图像融合是图像处理领域中的一个热门研究方向,其目的为了将来自不同传感器的多模态信息综合体现在一张高质量的图像上,已被广泛应用于医学、航空遥感、军事等领域。不同的融合算法会得到不同的融合结果,融合算法的选择直接决定融合的结果。方法:本文主要调研了当前比较热门的基于小波变换方法的融合方法。根据小波变换的流程,我们知道影响融合后图像质量的因素主要有两个:一个是变换分类及变换基,另一个是变换域的系数融合规则。本文将从这两方面对基于小波变换的各种融合方法进行总结。文中算法的选取原则为:融合实验效果好、被引用次数较多的文献中的使用的算法。另外,本文对经典的融合算法也进行了较系统的描述。结论:经过对文献的搜集与整理,我们就变换种类与融合规则方法分别进行了汇总:在变换种类上有传统Haar小波、性能经过提升的小波、与小波变换交叉使用的变换方法三个子类;在融合规则上,有单个像素法、区域法、多种决策算法参与的系数融合规则三个子类。最后本文叙述了几种对于融合后图像的图像质量评价指标。
Objective： Image Fusion is a hot spot in Image Processing field, which aims to combine different kinds of informa- tion from different sensors into an image with high informational density. It is always applied in Medicine, aerial remote sensing, military field, etc. The choice of algorithm has great influence on the fusion result. Methods： Select the mainstream Wavelet Transformation based fusion method as our fusion algorithm. According to the flow chart of Wavelet Transformation, the two main factors that affect the quality of fused image are： transtbrmation base and fusion rule of coefficients in the wavelet domain. This article will conclude fusion methods according to these two parts. Algorithms gathered here are based on two rules： algo- rithm with better fusion result, higher recitation frequency and popularity. Plus, some of traditional fusion algorithms are con- cluded briefly here as well. Conclusions： With gathering and analyzing papers of image fusion, we rearrange them according to transformation method and fusion rule. For the transformation method, it has three subcategories： traditional Haar wavelet, im- proved wavelet and hybrid transformation method. For the fusion rule, three categories as well： pixel, region and hybrid algo- rithm. Then, several kinds of fused image quality assessments are provided, which will help us evaluate image quality subjec- tively.
出处
《中国医学物理学杂志》
CSCD
2013年第6期4530-4536,4584,共8页
Chinese Journal of Medical Physics
基金
国家自然科学基金61201067
上海市教委科研创新项目13YZ069
上海市高校青年教师资助计划slg11017
关键词
图像融合
多模态
小波变换
融合规则
图像质量评价
image fusion
multimodal
Wavelet Transformation
fusion rule, image quality assessment
