一种基于局域统计特征的影像融合方法  

A Locally Statistical Based Image Fusion Method of Multi-scale Remote Sensing Data

作　　者：尤淑撑[1] 刘顺喜[2] 

机构地区：[1]北京师范大学地理学与遥感科学学院,北京100075 [2]中国土地勘测规划院,北京100035

出　　处：《遥感信息》2007年第2期8-11,I0001,共5页Remote Sensing Information

摘　　要：通过将较低空间分辨率的多光谱影像和较高空间分辨率的全色波段影像融合产生兼具多光谱和高空间分辨率特征影像是提高资源环境监测精度的重要途径之一。本文提出了一种计算简单、光谱保真性好的影像融合算法。该方法根据影像局域统计特征,建立融合方程,以表征全色影像空间结构和多光谱影像光谱信息的统计量为约束,求解融合系数,实现影像融合。试验结果表明该方法光谱保真性好、适应性强,具有一定的应用潜力。Image fusion is an important way to improve the accuracy of resource survey result. There are many image fusion methods, however, few of them are spectral reserving. In this article, we proposed a simple yet effective method, which is also spectral reserving. The method is based on locally statistical signature of panchromatic, multi-spectral and result images. Under presumption that the spectral information of result image comes from multi-spectral data and spatial structure information comes from panchromatic data, and both can be modeled locally by statistical method. Finally, we select Harbin as test bed and compare this method with IHS, PC and wavelet-based algorithms. The result shows that the method we provided is more effective than those methods mentioned above both on qualitative and quantitative aspects.

关 键 词：影像融合 统计分析 光谱保持 遥感 高空间分辨率 

分 类 号：P237.3[天文地球—摄影测量与遥感]