独立分量分析的图像融合算法  

Image fusion algorithm based on independent component analysis

作　　者：陈蜜[1] 伭剑辉[2] 李德仁[3] 秦前清[3] 贾永红[4] 

机构地区：[1]首都师范大学教育技术系,北京100037 [2]武汉大学计算机学院,湖北武汉430079 [3]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079 [4]武汉大学遥感信息工程学院,湖北武汉430079

出　　处：《光电工程》2007年第6期82-87,共6页Opto-Electronic Engineering

基　　金：国家自然科学基金资助项目(40204008);绘科技项目(1469990624201)

摘　　要：独立分量分析可实现图像的稀疏编码并具有能很好地捕捉图像重要边缘信息的特性。本文提出一种基于独立分量分析的图像融合算法,结合支持向量机对多聚焦图像的清晰域、模糊域进行判断以及在ICA域中进行图像分割以提取图像的主要边缘特征信息来实现特征级的多聚焦图像的融合。实验结果表明,本文提出的融合算法是有效的。Independent Component Analysis （ICA） was a recently developed linear data analysis method, which could realize sparse coding of images and capture the essential edge structures of the image data. A multi-focus image fusion algorithm was proposed based on ICA and Support Vector Machines （SVMs）. Using features extracted from the ICA domain coefficients, the SVMs were trained to classify the multi-focus images into clear regions and blur regions, and the images were segmented to extract the main edge information. Finally using the feature based fusion rules the corresponding clear regions of ICA domain coefficients and the edge information were fused into the composite representation. Experimental results show that the proposed algorithm is effective.

关 键 词：图像融合 独立分量分析 支持向量机 图像分割 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]