基于区域相关性的高分辨率遥感图像融合算法  

An Image Fusion Method Based on Regional Correlation for High Resolution Remote Sensing Images

作　　者：张凯[1,2,3] 李绪志[3] 

机构地区：[1]中国科学院光电研究院有效载荷应用中心,北京100094 [2]中国科学院大学,北京100049 [3]中国科学院空间应用工程与技术中心有效载荷应用中心,北京100094

出　　处：《计算机辅助设计与图形学学报》2014年第6期889-895,共7页Journal of Computer-Aided Design & Computer Graphics

基　　金：载人航天工程天宫一号民用遥感试应用项目(Y1141401SN)

摘　　要：针对基于现有的融合框架模型的遥感图像融合算法难以较好地同时保持融合结果的光谱特性和空间分辨率的问题,在分量替换融合框架模型的基础上,提出一种基于图像区域相关性的高分辨率遥感图像融合算法.该算法利用改进的脉冲耦合神经网络模型,根据全色图像和多光谱图像间的局部相关性差异对各多光谱图像进行区域分割,根据区域的相关性的差异确定相应的融合规则.实验结果表明,文中算法能取得很好的融合效果.State-of-the-art remote sensing image fusion methods based on component-substitution framework model have difficulty to preserve the spectrum characteristics and spatial resolution at the same time.In this paper,a remote sensing image fusion method based on regional correlation of the images is proposed.A novel pulse coupled neural network is proposed for image segmentation based on the difference of local correlation between the panchromatic image and the multispectral image,while the fusion rules are determined by the differences in regional correlation.The experimental results show that the proposed method performs very well for the remote sensing image fusion.

关 键 词：分量替换 高分辨率遥感图像 脉冲耦合神经网络 图像融合 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]